That’s when we had the first call that led us to build this newsletter.

We started The Workflow because the AI hype was killing us.

Every week, a new model launch. Every day, a new tool promising to replace marketers, designers, humanity. Every hour, a new LinkedIn post telling you that if you weren’t vibecoding your own SaaS, you were already behind.

Our first issue declared war on AI FOMO: this newsletter is for people who want to use AI to work smarter, not for the sake of using AI.

May 2026.

The noise has 70X-ed.

Claude Opus 4.6. Claude Sonnet 4.6. Claude Code. Gemini 3.1 Pro. Grok 4.20. GPT-5.4. Llama 4. Claude Opus 4.7. Claude Design. Claude Cowork. ChatGPT 5-5.

Product releases are faster than anyone can possibly adopt them. Social media is drowning in AI slop. For the first time, there is more supply than demand, both for products and for content.

So what did we do to keep this newsletter from disappearing into the content slop?

We promised you workflows that were easy enough to implement right after reading but challenging enough to give you new ideas, without ever omitting where AI falls short.

13 issues later, almost 2,000 people trust us with their inbox. Some even paid for a subscription without us asking.

Here’s everything we’ve shared so far:

🔥 The AI Website Audit Workflow roasting homepages in minutes

📚 The Marketing Report Workflow to slash production time by 50%

💡 The LinkedIn Content Workflow for scaling thought leadership

🗣️ The Voice of Customer Workflow to power your content strategy

🤝 The Buyer Validation Workflow to align Marketing and Sales

🤖 The BOFU Workflow for killer product-led content

🧮 The AI Lead Scoring Workflow to 2x free-to-paid conversions

🏅 The Brand Positioning Workflow for the AI era

💛 The Harmony Workflow to scale evergreen content

🤖 The Brand Mention Workflow to fix what LLMs say about you

🚀 The AI Workflow scaling 1600+ comparison pages with Clay

🤘 The Claude Code Workflow that enforces your brand style guide

😴 The Content Repurposing Workflow that runs while you sleep

These issues are not built to give you FOMO, and they’re not written by people who screenshotted n8n nodes to rack up likes on social. They’re built to help you think and work better. That’s the only promise we’ll always keep.

Subscribe now

What’s next 🔮

We’re taking a short break to sharpen our newsletter format.

When we come back in June: more solo issues, where we share the workflows running behind Sara’s advisory work, behind Diandra’s agency, and behind our own content systems. More guests too: CMOs, builders, and tons of smart people doing the lord’s work without getting the visibility they deserve.

The Workflow is free for now. Considering the time we both put into it, we’re exploring ways to keep it sustainable, through the right sponsors and, eventually, a paid tier. Some readers have already chosen to pay before we made it mandatory, and those are the people who will lock in the best price we’ll ever offer.

Dig into the archive while we’re gone. The beauty of strategy over hype is that no matter how fast models change, the logic behind these workflows holds.

We’ll be back in June with new workflows worth stealing.

Sara Stella & Diandra✌🏼

Behind every “consistent creator” is either a team, a burnout waiting to happen, or a system nobody’s sharing. For the past few years, the default advice has been “repurpose your content,” which is technically correct but practically useless without the infrastructure to back it up.

This week’s guest is Brandon Smithwrick, a content marketing director turned full-time creator who built his audience across roles at Ralph Lauren, Squarespace, and Kickstarter before stepping out on his own in August 2025. Brandon writes Content to Commas, a weekly newsletter with 8K subscribers, has 55K+ followers on LinkedIn, and holds brand partnerships with Canva, Zapier, and HubSpot. The system he shares today turns a single newsletter issue into a full week of platform-specific content across all six platforms he’s on, automatically, and without the output sounding like it was written by a robot who just discovered em dashes.

Subscribe now

✅ What you’ll learn

• How to use AI and automation to build a content repurposing workflow

• How to turn long-form content into platform-optimized posts for 6+ channels

• How to set up a never-ending content pipeline that keeps running while you’re asleep

🧨 What triggered The Workflow

This workflow exists to solve a pain every creator and content manager knows all too well: the need to stop working at 3am.

For months, Brandon was running his day job as Director of Content & Creative at Kickstarter while writing a newsletter and posting across six platforms. The content was good. The pace was *BONKERS*. “I felt like I had three full-time jobs,” he said. “And I wanted to still be very intentional about my content.”

Brandon knew that copy-pasting a LinkedIn post into Threads, or turning a newsletter section into a LinkedIn post without rewriting it, was what he calls attention leakage. You “repurpose” content, but it falls flat because it was clearly made for a different channel. He needed a way to take one piece of deep, long-form writing and turn it into many, without losing the platform-specific voice that made each worth consuming.

🤯

🛠️ How to build The Workflow

Step 1: Define your content home base

Everything starts with identifying your single source of content truth. For Brandon, that's his weekly Beehiiv newsletter. For a podcaster, it could be a transcript. For a founder, maybe a recorded interview. The format doesn't matter as much as the substance: one piece of long-form, in-depth writing that feeds the entire content waterfall.

Step 2: Build your monster master prompt

This is where most people skip ahead, then pay for it later. Brandon spent quite some time perfecting a prompt that defines two roles (senior content strategist and viral social media expert), his tone of voice, his audience by platform, and his goals per platform. The prompt also includes hard writing no gos: no em dashes, no clichés, no AI-sounding filler. It's less a prompt and more a content brief drafted with ChatGPT’s help.

👉🏼 Grab Brandon’s Content Waterfall OS prompt here before continuing to the next step.

Step 3: Set up the Zap in Zapier

The workflow trigger is the newsletter going live. In Brandon’s case, when a new issue is published on Beehiiv, Zapier fires and creates a Google Doc that captures the publish date, the link, and the full content of the issue. From there, Zapier runs two sequential AI steps: first, a synthesis prompt that distills the newsletter into a source document (core problem, target audience, thesis); second, the full Content Waterfall OS prompt (👉🏼 grab it here) that generates platform-specific content for pretty much any social media network from that synthesis. Both prompts live inside the Zap itself.

Once the AI steps are complete, Zapier creates a second Google Doc where all the platform-ready content lands, ready for review. Zapier lets you swap between models, but Claude Sonnet is your best bet for this job.

Step 4: Human in the loop: review the output

The output doc includes three short-form video scripts (with three hook options each), five short LinkedIn posts, one long-form LinkedIn post, twelve short text posts for Threads, five Twitter/X posts, one Instagram carousel outline, one newsletter draft, and one YouTube long-form video script. That's 29 pieces of content from a single newsletter issue. Brandon goes through it every Monday morning for about an hour: approving, trimming, rewriting where needed. "I still want to be the author," he said. "AI handles the translation. I handle the final call." This is the 50/50 line true creators won't cross.

Step 5: Schedule the content

Approved Threads and LinkedIn posts go into Sprout Social's Smart Queue. Brandon doesn't set individual publish times; he sets a daily cadence (3 daily Threads posts and 2 daily LinkedIn posts, and 4-5 times weekly for every other platform) and adds content to the queue as he approves it. Sprout handles the rest. For LinkedIn specifically, he either publishes manually or has his VA post on his behalf. Regardless of who posts, the content stays platform-native.

Step 6: Publish to Substack and seed Substack notes via WriteStack

Brandon uses Substack as a traffic channel, not a content home. He publishes around 50% of his newsletter issues there, always partial, with a clear pointer to Beehiiv for the full version. On top of that, WriteStack auto-publishes short-form posts to Substack Notes continuously, keeping him top-of-mind in the feed without any manual effort. "For the full article, subscribe to my Beehiiv. That's the tactic." With 8K subscribers in his first year, the numbers back him up.

Step 7: Batch engagement

Automating content creation is the easy part. The bottleneck most creators hit next is engagement: six platforms, dozens of comments, and the human need to sleep. Brandon’s answer is to batch engagement and be strategic about the “how”, rather than focus on sheer quantity.

On LinkedIn, he lets a post run for 12 hours before replying to comments. The algorithm gets time to distribute, and his engagement spike re-ignites reach at exactly the right moment. On Threads, he uses Wispr Flow to voice-dictate replies in 15-minute sessions. Instagram and TikTok get attention whenever he logs in, but he’s not chasing real-time interaction on every platform. “I’m going to let the algorithm do its job”, he said. The goal isn’t to be everywhere at once. It’s to show up with enough consistency that the platforms keep working in his favor.

🤖 Tools powering the Workflow

This is one of those workflows where a minimal tech stack compounds into an outsized output. You could replicate the automation layer in Make or n8n, but for non-techies, Zapier remains the most intuitive automation platform, especially now that it supports agentic workflows in plain language. No developers, no API wrangling, and seriously…no excuse not to build it!

🎢 Highs, lows and Workflow warnings

✅ What shines

• One input, 29 pieces of content

  A single newsletter issue generates a full week of platform-ready content across six channels. Brandon multiplied his content output and reach without writing anything new on top of his newsletter.

• No technical barrier to entry
  Zapier’s plain-language builder means anyone can set this up. If you can describe what you want to automate, you can build this Zap.

• The system gives you back your time
  The content waterfall keeps flowing while you're at the gym, asleep, or eating ramen in Tokyo.

⚠️ Workflow warning

• Underinvest in the master prompt, and the whole system falls apart
  The Zap takes five minutes. The Content Waterfall OS prompt requires an extensive library of high-quality content per platform and a solid content strategy to draw from: your tone, your audience per platform, your goals, your non-negotiables. If you skip that work, the output will sound like everyone else’s AI content, technically correct, robotic and personality-free.

• The human in the loop > the automation
  Reviewing the output doc, approving, trimming, and rewriting where needed will still take a couple of hours per week. Brandon publishes most posts manually or through a VA, and for video, Sprout sends a push notification to his phone so he can publish natively rather than auto-scheduling. “I want to work with my AI workflows. I don’t want my AI workflows to do every single thing for me.” Automation handles the volume. The creative call stays yours. What’s left of a creator if you automate away the act of creating?

✨ The Goldflow

Anyone can prompt an AI to turn a LinkedIn post into a YouTube script. What makes the difference is the brainwork you put in before the automation runs.

“Subscribers aren’t the goal, that’s going to be an outcome. What’s your goal with YouTube? It’s to build trust and to make complex things simple. Done, that’s your goal.” That level of clarity, per platform, is what separates your content waterfall from an automated, cross-channel copy-paste.

Brandon’s clothing analogy stuck with us long after the interview. “LinkedIn is an Oxford shirt. Instagram is a polo. TikTok is a tank top. Same person walking into every room, but the outfit sets the tone for the conversation.”

Turns out, you can’t systematize a voice you haven’t defined yet (and you better choose your outfit intentionally).

Sara Stella & Diandra ✌🏼

👔 Send this to that person who's copy-pasting LinkedIn posts into Threads and calls it repurposing.

Share

😢 Can’t survive two weeks without us?

Catch up on past Workflows and steal the playbooks you missed.

• 🏅 The Brand Positioning Workflow for the AI era

• 🤖 The Brand Mention Workflow to fix what LLMs say about you

• 💛 The Harmony Workflow to scale evergreen content

Subscribe now

Welcome back to The Workflow, where we think AI hype is a new kind of yellow fever.

Unless you live under a rock, you must have seen the Claude Code posts. “Marketers who aren’t using this will be obsolete by Q3.” “I automated my whole life, and I’ve never felt more alive.” We figured we owed it to you to check what’s noise and what’s real.

So we called Tim Metz, Director of Marketing and Innovation at Animalz, one of the most respected B2B SaaS content marketing agencies in the space. Tim has a basic coding background and the specific kind of brain that, once it discovered Claude Code, immediately started automating everything in sight: content workflows, LinkedIn post generation, AEO audits, and yes, at some point, his own sleep tracking. The terminal is now his natural habitat.

His workflow for this issue fixes a problem every content team has and rarely admits to: your brand style guide exists in a Notion doc, gets mentioned during onboarding, and enforces absolutely nothing in practice.

Tim’s Claude Code workflow reads your published articles, extracts your actual editorial patterns, turns them into a working brand style guide, and then runs eight parallel agents to check any new article against it, line by line. He built the whole thing in about 20 to 30 minutes, which means by the time you finish this issue, you could have one running too.

Subscribe now

✅ What you’ll learn

• What Claude Code actually is and how it works

• How to generate a brand style guide from your existing published content, without manual work

• How to run eight parallel AI agents to copy-edit any article against your brand style guide

• How to set up Claude Code even if the word “terminal” makes you cry

🧨 What triggered The Workflow

At Animalz, content goes through a meticulous editorial process: writer to editor to copy editor, with each step significantly impacting quality and stylistic nuance. For the agency’s own marketing, the same rules apply… in theory. Client work always takes front row, and Tim runs internal marketing largely solo, which means when a newsletter issue goes out on a Thursday and a blog post needs to be ready by “yesterday”, his editor is often staffed on external projects.

“Unfortunately, it’s happened that I published an article on the Animalz blog without spending enough time on the style editing review,” he told us. Anyone who has ever been a team of one knows exactly how that happens.

The second problem: because some articles skipped the process, the brand style guide had drifted. Certain rules were respected in some articles, but ignored in others. Nobody had a reliable picture of what the actual house style even was anymore. Brand style guide maintenance is the kind of task that always gets added to the list but never makes it to the top.

Tim had tried solving the problem with AI before. ChatGPT, earlier Claude models, various prompting approaches. They would catch some things and miss others. Inconsistently. Unpredictably. The problem was much bigger than the model: one context window, one pass, too much context and too many assets to hold at once.

Claude Code changed the constraint. And that changed the game for internal marketing at Animalz.

🛠️ How to build The Workflow

Before diving in: Claude Code runs in your computer’s terminal, which sounds more intimidating than it is. You don’t need to write code or trash your English literature degree. You can navigate it entirely in plain English, just as you’d use Claude, ChatGPT, or talk to us. The setup takes one command and about five minutes. If you want the full installation walkthrough, Tim wrote a no-code guide for content marketers on the Animalz blog.

👉🏻 Read it here

You’ll need a paid Claude account (Pro or Max) to run Claude Code.

Step 1: Set up your project folder

Create a dedicated folder on your computer. This is where Claude Code will pull documents from and upload newly generated ones to. Call it something like blog-style-guide-creator so Claude Code immediately understands the context from the folder name alone. File naming and folder structure are not just organizational habits here: they’re part of how Claude Code understands what it’s working with. Make sure to be laser-sharp!

Inside this folder, you’ll need two types of files:

1. The command files tell Claude Code what to do when you run a slash command. There are two: one for /generate-style-guide and one for /style-check. Each is a markdown file containing a detailed prompt that describes the full process Claude Code should follow. No need to write these from scratch.

2. The agent files are what make the style check work. There are eight of them (!!!), one per editorial dimension. Each defines the specific agent’s role, what it’s checking for, and how it should format its output. Claude Code reads these when it spins up the parallel agents we’ll see in Step 6.

👉🏻 [Download the full zip folder here]

⚠️ NOTE: No need to write any of these files from scratch. The templates in the zip are ready to use. Download, swap in your brand name, tweak for your own content, and run.

Once you’ve placed all the files in the folder, open your Finder, right-click the folder, and select “New Terminal at Folder.” Type claude and hit return. Claude Code will ask if you trust the files in this folder. Say yes, because you do. You’re in!

Step 2: Collect your source articles

Before you generate anything, Claude Code needs material to learn from. These are your published articles, the ones that already represent how you write at your best.

You have two options for providing them. Either drop local markdown or text files directly into the folder, or have a list of URLs ready to paste when Claude Code asks. If your blog has a sitemap (most do, usually at yoursite.com/sitemap.xml), you can point Claude Code there, and it will find the articles itself.

Tim recommends 10 to 15 articles for best results. Make sure to choose pieces that represent your typical writing, not your one-off content experiments. The brand style guide gets built from whatever you feed it, so a skewed sample will produce skewed rules.

Step 3: Generate your brand style guide

Type /generate-style-guide [your brand name] and hit return. Claude Code will ask how you want to provide your source articles: a blog URL to scrape, specific URLs to paste, local files from your folder, or a mix. Pick whatever matches how you collected your articles in Step 2.

From here, Claude Code takes over. It fetches and stores the full text of each article locally and runs a brand style guide generator agent that analyzes patterns across eight editorial dimensions. This is the part that would take a human HOURS (and too much mental sanity) to do manually and still be full of errors.

The output is a markdown file saved inside your brands folder, with 50+ specific rules drawn directly from how you actually write. The whole process takes a few minutes.

👉🏻 See the example brand style guide we generated for The Workflow

Step 4: Review the AI-generated brand style guide

Open the markdown file in your brands folder and read through it. Claude Code extracts patterns from your writing accurately most of the time, but it will occasionally give you false negatives or propose rules that don’t reflect your style. A brand style guide made of wrong rules enforces the wrong logic every time you run it, so catching errors here saves you from compounding them later. It’s AI, friend, not magic. ✨

Add missing rules, remove the ones that feel off, and make sure the language is yours. The guide is an easy-to-edit .md file.

Step 5: Run the brand style check on any article

Type /style-check [brand name] [article] and point it at a local file or URL. This is where the architecture does what a standard chat interface cannot.

Claude Code spins up eight parallel agents simultaneously (!!!), one for each section of the brand style guide. Each agent has a single focus and its own context window: one looks exclusively at voice and tone, one at grammar and usage, one at punctuation, and so on. Because they run in parallel rather than sequentially, the full check takes 5 to 10 minutes. A manual review covering the same eight dimensions would take 40 minutes or more.

The reason this catches what a single-prompt AI review misses is structural. One context window doing eight jobs at once will always deprioritize at least a few of them. Eight agents, each doing one focused job, do not.

Step 6: Apply the corrections

Claude Code outputs a detailed report flagging every violation it found, with the line number, the exact text, the rule it violates, and the suggested correction. Not “consider revising this section.” More like: line 47, passive voice, here’s the active version.

Tim ran it on a published Workflow article during our session, and it caught a metric claim without attribution, a poorly formatted structure, and a missing overview section in a piece that had already gone through editorial Sara’s review (Oooopsie).

Step 6: Update and finalize the article

We got three options for you. Fix issues manually by working through the report yourself, which gives you full control over every change. Run --auto-apply to let Claude Code make all corrections automatically and save a new version of the file. Or run --dry-run first to preview every change before committing.

For a first pass, --dry-run is worth it. --auto-apply on a piece you haven’t read carefully is where things get sloppy. The final, corrected article is saved as a new .md file in your folder, ready for publication.

🤖 Tools powering the Workflow

Kidding, there’s only ONE. Claude Code is all you need to run this entire workflow, which is either a relief or mildly suspicious, depending on how you feel about terminals.

If you want to go deeper on Claude Code beyond what this issue covers, here are two of the best no-hype guides we’ve come across (on top of Tim’s, of course!):

👉🏻 Claude Code for Everyone by Carl Vellotti

A free course taught entirely inside Claude Code itself. No videos, no docs, no coding experience required. You learn by doing real work inside the tool, guided by Claude at every step. The fastest way to go from zero to actually useful.

👉🏻 Claude Code for Everything by Hannah Stulberg

A five-part series that covers setup, core workflow, and a genuinely useful explainer on why AI gets worse the longer you talk to it, and how Claude Code fixes that. Written for non-technical people who want to use it for real work, not just coding.

🎢 Highs, lows and Workflow warnings

✅ What shines

• It catches what everything else misses

  The parallel agent architecture is not a gimmick. A single-prompt AI review runs one pass and hopes for the best. Eight focused agents running simultaneously, each with its own context window and a single editorial job, produce a qualitatively different result.

• The style guide builds itself from your real writing
  Most style guides get written once from scratch, live in Notion, and slowly become redundant. This one is extracted from how you actually write, which means it reflects your actual patterns rather than your intentions. That’s a HUGE difference.

• The setup cost is low relative to the ROI
  Once the folder is set up and the style guide is generated, running a style check on any new article only takes a single command.

❌ What doesn’t shine

• Token usage adds up fast
  Eight parallel agents running simultaneously eat up significantly more tokens than a standard chat interaction. If you’re on a Pro plan and running this frequently, keep an eye on usage. You’ll Max it out fast (pun intended!).

• Watch your instructions
  During our session, Tim caught the style guide generator using article summaries instead of full text. Not a hallucination: the prompt simply hadn’t specified full articles, so Claude Code defaulted to summaries. A small ambiguity in the instruction, a HUGE difference in the output.

⚠️ Workflow warning

The biggest risk here is behavioral. "It's so tempting to just hit next on what it proposes if it looks decent." Touché. A workflow that runs fast and spits out something that looks polished creates the perfect conditions for sloppy oversight. The style guide review in Step 4 and the violations report in Step 6 both require a human who is actually paying attention, not one who is relieved the hard part is over.

✨ The Goldflow

Tim literally built this workflow from a single voice memo. One recorded description of the problem, and Claude Code created 90% of the rest autonomously. 20 to 30 minutes, start to finish.

Tim’s also the first to flag what that speed costs you. “It’s surprisingly easy to go from being detail- and craft-oriented to lazy, letting AI take over and falling asleep at the wheel.” A workflow that builds itself in half an hour also lowers your standards in half an hour.

Tim is convinced AI will soon replicate anyone’s writing style perfectly. The models are almost there, and context windows are growing fast enough to hold the full spectrum of what makes your voice yours. But he doesn’t think that’s a reason to stop writing. “I haven’t found any alternative for working through your thoughts to clarify an idea that comes even close to the process of writing.”

So, friend, you can automate the style check. You can’t automate the thinking that makes the writing worth checking.

Sara Stella & Diandra ✌🏼

🤝 Forward this to someone if you’re dead tired of the Claude Code hype, but for whatever reason still love this article

Share

🌯 It’s a major wrap

This was The Workflow, a bi-weekly newsletter by Sara Stella Lattanzio &Diandra Escobar, featuring one real workflow that moved the needle for someone even smarter than the system they built.

Subscribe now

This week’s guest is Matteo Tittarelli, founder & solopreneur at Genesys Growth, a B2B SaaS product marketing and content consultant whose AI workflows regularly make us stop scrolling. If you’ve seen him document his work on LinkedIn, you’ll recognize him as a true GTM engineer in action.

Matteo is a solo operator, which usually means capacity is the ceiling. With AI, that ceiling stretched to the max, allowing him to scale his output without scaling headcount.

In 2025, that leverage allowed him to handle up to 6 clients per month on his own, while keeping a (mostly) normal workweek without turning his calendar into a sleep-deprivation experiment.

Today, we scrutinize one of his wildest and most successful AI experiments so far. Scaling competitor and category comparison pages (1600+, to be exact), killing manual work while keeping the quality bar high.

Impossible? You’ll see.

Subscribe now

✅ What you’ll learn

• How to scale 1600+ competitor pages without sacrificing quality

• How to scale SEO content without producing AI slop

• How to use Clay for programmatic SEO

🧨 What triggered The Workflow

Comparison pages became a recurring client request for Matteo. SaaS in crowded markets with thousands of products and features need content that explains why they’re different (and better). There are so many ways to approach this: comparing tools, feature categories. Then adjacent categories. Before long, you’re staring at hundreds of pages.

Doing this manually wasn’t an option if he wanted to keep his sanity. Pricing changes. Reviews age. Feature claims drift. At scale, errors spread fast.

So the GTM engineer re-engineered the process.

He built a workflow that pulls together live market data, reviews, and pricing, anchored by a structured positioning library. The result is a scalable comparison-page engine that stays fresh, doesn’t become a maintenance nightmare, and drove a 38% increase in organic impressions just two weeks after the first pages went live.

🛠️ How to build The Workflow

To build this workflow, start by defining what matters in your ICP’s buying decision. Only then use AI to handle the grunt work needed to scale. You don’t just want to create pages, you want to express how you see the market.

Here’s how the pieces connect, from the positioning that drives every comparison to a scalable hub of product marketing pages.

Steal Matteo’s prompts and save yourself the trial and error.

Step 1: Build a structured messaging and competitor library

Ahhhhh, the good old foundational work required to do anything interesting with AI.

Use Octave to centralize the building blocks of a company’s narrative. Core messaging, sales decks, customer intelligence, and selected competitor data live in one structured place.

Each product is broken down into the same core blocks: description, differentiated value, capabilities, problems solved, and customer proof. That structure also applies to personas, use cases, segments, reference customers, and competitors.

“You’d be surprised how many companies don’t have a real source of truth for their messaging.”

Matteo, we believe you.

Congrats! You just created a reusable positioning library that feeds comparison pages, sales enablement, and outbound personalization, without breaking consistency.

Step 2: Messaging fine-tuning

The last thing you want is to scale wrong or inconsistent messaging.

Use Octave’s natural language chat interface to review what was ingested, correct inaccuracies, resolve contradictions, and clean up inconsistencies across products, personas, and competitors in your library.

Step 3: Wireframing

While refining the messaging, it’s critical to think ahead about how that information will live on the pages. This step intentionally overlaps with messaging work to prevent bottlenecks later.

Matteo recommends collecting examples of ideal landing pages, then building section-by-section wireframes for each page type in Figma. Make sure to follow a clear hierarchy:

• a central competitor hub that acts like a homepage,

• reusable templates for product vs competitor, product vs category, and competitor vs competitor pages.

Each template should follow the same core structure, from navigation and hero to comparison tables and supporting sections.

Every wireframe is designed with future automation in mind. Each section corresponds to a specific data input that will later be generated and populated programmatically. Locking the structure and hierarchy at this stage allows the system to scale to hundreds of interlinked pages without collapsing later.

Step 4: Build the programmatic comparison powerhouse in Clay

With the wireframes locked, move into production. Clay is the workspace where every comparison page is assembled at scale, using the messaging and positioning already defined in Octave.

Instead of creating pages one by one, Matteo creates tables. Each table represents one type of page. One table powers product vs competitor pages. Another powers product vs category pages. A third handles competitor vs competitor pages for any specific category (and you can scale this for as many adjacent categories you have!). The layout of each table is almost fixed (only a few comparison fields can change, but it’s up to you). What changes from row to row is the page’s subject, such as a specific competitor, category, or pairing.

Think of each row as one future page, and each column as one piece of that page. That’s the foundation everything else builds on.

First come the input columns. These define what each row represents, such as a competitor, category, or URL. This is the minimum information needed to anchor a page.

Next are the research columns. Clay scrapes competitor websites, pulls reviews from sources like G2 and Capterra, and fetches logos via Brandfetch.

Then come the copywriting columns. This is where Octave comes back in. Use the Octave content agent inside Clay to pull messaging and positioning from the library created in step one. That context is combined with the research data to generate SEO metadata, headlines, comparison tables, pricing sections, differentiation paragraphs, and FAQs.

Because each column maps to a specific section of the page, Clay is able to fill predefined slots.

Tip: Work in small batches to protect quality and avoid burning credits (or connect your Anthropic API for saving Clay credits altogether!). Test prompts on single rows, refine them, and run the full table only then. Once it holds, the setup can be reused across categories with minimal changes.

Grab Matteo’s best prompts to populate your Clay tables

Step 5: Push pages live to your Content Management System

Time to bring these gems live! Matteo loves Webflow as a CMS because it offers precise field mapping and solid SEO controls.

Each output column in Clay maps to a specific Webflow CMS field. Headlines, subheadlines, logos, comparison tables, pricing sections, FAQs, and SEO metadata all have predefined destinations, which makes publishing smooth as butter.

Pages can be pushed either to staging or directly live, but one rule holds: never publish too many at once. Start with a small batch, review it carefully, fix copy or layout glitches, and only then scale up to publishing hundreds of pages.

The real win is maintenance. Because pages are generated from Clay tables, updates don’t require manual rework. Dynamic fields like pricing or reviews can be rerun and pushed to Webflow without turning updates into a nightmare.

🤖 Tools powering the Workflow

This isn’t an easy workflow, but the logic applies to any project that needs to scale. AI and automation only amplify the foundation you give them. Here, positioning comes first and speed follows. And yes, Matteo more or less invented a whole new use case for Clay 🤯

🎢 Highs, lows and Workflow warnings

✅ What shines

• Crazy results, fast
  It took about 1.5 months to launch the first 500 pages. But two weeks after launch, organic impressions were up ~40%, clicks rose ~150%, and CTR doubled (on an already three-digits baseline ;). Fresh pages also started surfacing regularly in AI overviews for queries like “[competitor] pricing”.

• Set up once, reuse forever
  The real strength of this workflow is its versatility. The same structure supports every type of comparison page, so expanding into new categories like “[your brand] vs [category]”, “[competitor] vs [category]”, or “[category] vs [category]”, or adding new players, becomes almost plug-and-play.

❌ What doesn’t shine

• Quality assurance is a pain
  Even when everything looks perfect in Clay, issues tend to pop up once content hits the CMS. Fields don’t always stitch together cleanly (especially with custom HTML code blocks), mappings break, and small glitches appear. Expect manual fixes and ongoing QA maintenance, especially at scale!

• Slow time to value
  The upfront work is heavy. Wireframing, design, prompt tuning, tone calibration, CMS field mapping, and custom HTML code and workflow setup all take time. Matteo spent roughly two months getting the first large batch live. After that, the ROI compounds.

⚠️ Workflow warning

This workflow only makes sense in hypercompetitive, mature markets with complex, feature-rich products. To make it work, you need to know how to build strong landing pages and understand both buyer intent and LLM scraping behaviour. AI and automation aren’t the only hard part here.

✨ The Goldflow

While everyone and their mother is obsessing over scaling faster, Matteo is obsessing over getting the fundamentals right.

“AI needs to be told what good looks like. That only works if you have context: your ICP, your market, your competitors.”

Yet another reminder that scale and tools aren’t strategy. They never were. They never will be.

Sara Stella & Diandra ✌🏼

🏗️ Send this to the friend who still builds comparison pages one by one.

Share

😢 Can’t survive two weeks without us?

Catch up on past Workflows and steal the playbooks you missed.

💡 The LinkedIn Content Workflow for scaling thought leadership

🤖 The BOFU Workflow for killer product-led content

🧮The Clay Lead Scoring Workflow

Subscribe now

LLMs care less and less about what brands say about themselves. How others describe them carries way more weight.

Oh, the irony. Teams keep shipping fresh content on autopilot while the internet keeps repeating a brand narrative from three positioning pivots ago. This quietly kills online visibility because, unlike SEO, AI search gives you no noindex button.

Today’s guest is only in his mid-20s, but he understands how online search works better than most people ever will. Usman Akram, Discoverability Lead at influencer marketing platform Modash, built a workflow to help brands take back control of their narrative in AI search. It audits how the internet understands your brand, highlights gaps against your intended positioning, and tells you exactly what to fix first.

Less “publish and pray.” More “make the web speak with your brand voice.”

Let’s get to work 👇🏼

Subscribe now

✅ What you’ll learn

• How to build an AI workflow to improve your brand positioning in AI search

• How to find third-party brand mentions across the web and compare them with your intended positioning

• How to prioritize which positioning mismatches to fix to improve your AI visibility

🧨 What triggered The Workflow

In a world where no one can say with 100% certainty how to show up in AI search (despite everyone having a hot take), Usman had a clear edge.

When he built this workflow, he was working at a well-known SEO agency in the middle of its shift into AEO. He was advising and managing dozens of client accounts, from fast-growing startups to brands with millions of existing mentions across the web.

That put him in a privileged position: deep SEO fundamentals, a ton of curiosity about how LLMs work, and enough real-world data to test his hypotheses instead of pontificating on LinkedIn.

One pattern kept showing up. Companies would evolve their positioning and expect a few fresh brand mentions to magically shape how AI talks about them.

“That’s when it clicked,” Usman told us. “LLMs care about recency, but they care even more about consensus. Adding X new mentions a month doesn’t change anything if the rest of the internet already spreads a different story.”

That’s when he flipped the approach. “Dear brand, before creating anything new, what if you fixed what already exists? What if you identified where LLMs get your brand wrong and corrected the highest-impact mentions first?”

Now he had to build something to make that possible. But not a content engine as he’d use for traditional SEO.

More like a brand guardian.

🛠️ How to build The Workflow

This workflow is essentially an internet-wide positioning audit. You feed it a brand name and a domain, and it will define

• how you describe yourself today

• scrape how the rest of the web describes you

• surface where those two stories don’t match, and

• tell you which mismatches you should fix ASAP.

In theory, you could stitch this together with n8n or any other workflow builder. In practice, Usman runs it in AirOps, which he half-jokingly describes as “n8n for SEOs.”

It’s built for people who understand the logic of automation but don’t want to wrestle with raw APIs all day. Native tool integrations remove much of the setup work, and for the more elaborate steps, Usman used AirOps’ copilot to generate code snippets without being a coder.

No panic, we’ll walk through each step. And yes, there will be prompts you can steal along the way, because we don’t do gatekeeping over here.

🤖 Get the prompts to replicate the workflow

Step 1: Input brand details

The first AirOps step collects two inputs: the brand name and the primary domain you want to examine.

Domains alone aren’t super reliable input, so make sure to add the full brand name as well. Later steps will rely on exact string matching to find the right brand mentions across the web.

Once this step runs, both values are passed forward as variables to all subsequent steps and reused throughout the entire workflow.

Step 2: Scrape the homepage (brand positioning blurb)

IMPORTANT: Make sure your homepage reflects the latest version of your company’s positioning. In this step, we scrape your homepage and pass the content to an AI step. The AI model (in this case ChatGPT 4.1) is instructed to produce a short, plain-language brand blurb that reflects how the company currently describes itself.

At this stage, the output is only a draft. The goal is to answer, as accurately as possible: “Based on this homepage, how would you describe this brand today?”

Before moving on, remember to proofread and edit the blurb if needed. Once it’s final, this version becomes the reference point for all later comparisons, so it should reflect the best version of your brand positioning.

👉🏼 Get the prompt to generate the brand positioning blurb here

Step 3: Find existing brand mentions across the web

With the positioning baseline locked, the workflow moves into discovery.

Using the brand name from Step 1, AirOps searches the web for third-party pages that mention it. For each result, it captures the URL, page title, and the snippet describing the brand. At this point, you will receive a list of websites that mention your brand.

Don’t worry about pulling too many results here. This step is meant to be broad, and we’ll prioritize what matters later.

Step 4: Score positioning consistency

For each brand mention found in the previous step, an AI Step compares the third-party snippet to your approved brand positioning blurb. The output is a positioning consistency score from 0 to 5, where a higher score means the page description and the brand blurb are fully aligned.

Remember: this isn’t sentiment analysis. A source can praise the brand and still get a low score if it frames the brand in the wrong category, uses outdated language, or surfaces old positioning.

At the end of this step, every mention has a clear signal attached to it: how closely it matches your current positioning.

👉🏼 Get the prompt to score URLs mentioning your brand here

Step 5: Enrich mentions with traffic and ranking data

Once each mention has a consistency score, the workflow adds context.

For every URL, it pulls keyword rankings and search volume using Semrush data, which is natively integrated into this AirOps step. This is a proxy for traffic potential and shows which third-party mentions matter the most for users, AI search and traditional SEO. By the end, each mention is weighted by how visible it is.

A misaligned mention on a high-traffic page is a much bigger problem (and a much bigger opportunity to fix) than a perfect mention no one will ever see.

Step 6: Combine and synthesize all data

Now you have it all. Brand mentions. Positioning scores. Traffic potential.

Time to connect all the dots. Using a lightweight aggregation step, the workflow merges URLs, snippets, consistency scores, and Semrush-derived metrics into a single dataset in Google Sheets. Each row represents one third-party mention, fully enriched and ready to be evaluated side by side.

This picture is clear. You can finally see which mentions are misaligned and influential, and which ones are misaligned but irrelevant.

From here on, it’s time to act on what will actually move the needle for your online visibility.

Step 7: Prioritize pages and start outreach

Use that juicy Google Sheet to sort brand mentions by positioning misalignment and traffic potential to see which pages are actively shaping how LLMs understand your brand.

Organize outreach starting with high-impact pages. Ask publishers, directories, and listicle owners to update how they describe your brand so the web starts telling the right story about you.

In case you forgot… 🤖 Get the prompts here

🤖 Tools powering the Workflow

AirOps is the only tool you need to replicate this workflow. If you’d rather build your own adventure with n8n or another workflow builder, you can stitch SEO tools and AI models together yourself.

🎢 Highs, lows and Workflow warnings

✅ What shines

• It makes AI visibility measurable

  With this workflow, you can see exactly how the web describes your brand today, where that description is wrong, and what pages influence how you show up in AI search.

• It prioritizes impact, not busywork

  Not all brand mentions are created equal. By layering positioning consistency with traffic data, the workflow tells you what fixes will move the needle, instead of sending you on a never-ending outreach scavenger hunt.

• Human in the loop where it matters

  AI does the crawling, comparing, and scoring. Humans decide what “correct positioning” looks like and which changes are worth pursuing. That balance keeps the workflow useful instead of creating a to-do list for the sake of it.

❌ What doesn’t shine

• Large brands need tighter guardrails

  If your brand has thousands or millions of mentions across the web, the mention discovery step can surface outdated or low-quality pages. Traffic filters help, but additional manual qualification is still required.

• Outreach is still…outreach

  The workflow gets you to a clean, prioritized list. Getting people to change copy still involves emails, follow-ups, and some luck. Sadly, the workflow can’t hypnotize the internet (yet).

⚠️ Workflow warning

If your positioning blurb is vague, your scores will be vague. If your discovery step is too loose, you’ll get noise. And if you don’t sanity-check the output, you’ll prioritize the wrong pages with great confidence.

This workflow shines when you’re precise about how you define your positioning and disciplined about what you act on.

✨ The Goldflow

Usman is unapologetically bullish on AI because “It makes work possible that simply would not happen otherwise.”

This workflow is a prime example of that shift. Without AI and automation, attempting it would have been *bonkers* at best. As Usman explained, “Anything that can be delegated can usually be automated to some extent. Not fully and not blindly, but enough to meaningfully change how we spend time and attention.”

Fully automated systems often create false confidence, where neat scores and clean outputs push teams to act on the wrong things.

The sweet spot is automating just enough that good strategy becomes unavoidable.

So yeah, AI won’t replace you, but it’s making the internet’s opinion about your brand impossible to ignore.

See you in two weeks with another workflow that exposes poor strategy, whether you’re ready or not.

Sara Stella & Diandra ✌🏼

Can’t survive two weeks without us? Catch up on past Workflows and steal the playbooks you missed.

🔥 The AI Website Audit Workflow roasting homepages in minutes

📚 The Marketing Report Workflow to slash production time by 50%

💡 The LinkedIn Content Workflow for scaling thought leadership

🔧 Help us fix more confused brand narratives!

Share

🌯 It’s a major wrap

This was the eleventh edition of The Workflow, a bi-weekly newsletter by Sara Lattanzio & Diandra Escobar, where we unpack the playbooks that actually move the needle for marketers smarter than their tools.

Subscribe now

In our last issue, we unpacked Harmony, the sophisticated editorial engine powering Search Engine Land, a full-blown n8n orchestra built by tech specialists and serving an entire editorial team. Today we are sticking with n8n, but our guest brings a workflow that is definitely lower-maintenance.

Aldo Gnocchi, Program Director and Lecturer at Swiss University FHNW and Owner of the Senior Digital Expert Collective Gnocchi & Friends, has been experimenting on the internet since you had to pay by the minute and pray your new hobby didn’t bankrupt you. Today, he still gets the same spark in his eyes when he talks about what AI makes possible.

One of the biggest shifts for him has been cutting out the tasks that eat time but add little value. Creating website and homepage audits from scratch was right at the top of that list. His solution is a surprisingly simple n8n workflow that scrapes any webpage, analyzes it with an AI agent, and spits out a structured audit ready for a touch of human expertise.

It’s fast. It’s rough around the edges. And it saves him from the kind of repetitive work that makes even the most seasoned marketers question their career choices.

Let’s get into it 👇🏼

Subscribe now

✅ What you’ll learn

• How to build an AI website audit workflow [complex and simple version]

• How to prompt AI to audit any webpage like a seasoned CRO expert

• Where human review fits in so your audit doesn’t turn into AI slop soup

🧨 What triggered The Workflow

In theory, auditing websites should be a strategic exercise. In reality, it often feels like dental work: necessary, repetitive, and quite painful.

Aldo has been doing website and conversion audits for years, repeating the same steps, spotting the same patterns, and spending hours on work that felt far more mechanical than creative.

Same steps, same checklist, different website.

→ Is the headline clear or trying to say five things at once?
→ Is the call to action specific enough?
→ Are the visuals supporting the message or just filling space?
→ Does the structure match the typical customer journey?
→ Is anything on the page actually building trust?

After years of doing website audits, Aldo could feel the creativity draining out of the process.

“That is why I wanted to automate it,” he told us. “It is interesting work, but it is repetitive.”

So he did what a marketer facing a lightbulb moment eventually does: he opened n8n at 7 a.m. and built an agent-powered workflow to do the first 80% of the audit for him.

The whole thing started as a fun experiment… until he realised he never wanted to go back to doing things the old way.

🛠️ How to build The Workflow

Aldo’s automation is a six-step agent-powered workflow that runs end-to-end inside n8n.

You enter a URL, hit submit, and the system scrapes the homepage, analyses it with a Conversion Rate Optimization (CRO) agent, formats the output, turns it into a client-ready PDF, and stores everything in your Google Drive. We focus on a homepage audit here, but the process is the same whether you are auditing a landing page or an entire website. You simply adjust the prompts.

Here’s how the whole thing works:

Step 1: Collect the URL via a simple n8n form

Everything starts with a lightweight n8n form where you paste the URL you want to audit. The moment you hit submit, the entire workflow fires automatically. Anyone on Aldo’s team can now run an audit without touching the backend or configuring anything. Paste, submit, done.

Step 2: Scrape the homepage

Once the form fires, n8n sends the homepage URL to a scraper that pulls the full HTML and turns it into clean text. This gives the agent a proper snapshot of the page and avoids the hallucinations and errors that often happen when you feed it copy-pasted fragments.

The output from this step becomes the data packet passed into the CRO agent, which we will now refer to as the “homepage roaster.” Aldo did not call it that, but he really should.

Step 3: Roast the homepage

This is where the workflow turns into a real audit engine. The scraped text flows into Aldo’s homepage roaster agent, powered by a structured prompt shaped by years of hands-on CRO teardown experience.

Grab Aldo’s homepage roaster agent prompt here 🤖

What should you consider when auditing a homepage?

• Clarity of messaging

• Relevance for the target audience

• Value proposition and its strength

• Call to action quality and specificity

• Trust and credibility signals

• Accessibility and usability

• Visuals and design

• Mobile optimization

This is Aldo’s go-to list. Think of it as a flexible audit backbone you can adjust to any website or business model.

To keep audits consistent and comparable across sites, he uses a scoring system from 1 to 10 for each factor, which then adds up to an overall score. Within each category, the agent evaluates specific sections of the page and produces actionable recommendations tied to the areas that need improvement.

To run the agent, Aldo uses OpenRouter, an API hub that gives him access to any AI model through a single API into n8n, making experimentation across models a breeze.

For this workflow, Aldo picked Gemini Flash. Not because it’s the most advanced model, but because it’s fast and cheap per token. “A CRO audit doesn’t need creative genius; it needs consistency, structure, and speed.” Using an advanced reasoning model here would be “an overkill that just burns money.”

Grab Aldo’s homepage roaster agent prompt here 🤖

Even if n8n is not your thing, you can plug this prompt into any LLM or adapt it to whatever AI tool you already use.

Right after this step, Aldo uses a small formatting agent to turn the output into clean markdown so the HTML node can process it. It is a quick workaround that saves him from writing custom code. A simple custom node could do the same job for free, which is why Aldo sees this solution as functional but a bit of an overkill.

Step 4: Convert the audit into HTML

Once the markdown is ready, n8n passes it into a standard conversion node that turns the structured text into HTML. This is the point where the audit stops being a block of text and becomes a structured document with headings, lists, and a clear hierarchy.

HTML is the ideal intermediate format because it keeps the structure consistent and prevents formatting errors that often appear when exporting directly from an LLM. Aldo reviews the output here, while it is still editable, before the next step where n8n renders the HTML as a clean, client-ready PDF.

Steps 5 & 6: Generate the PDF and store it in Google Drive

After the HTML is created, n8n passes it to a PDF generator node that renders the audit into a clean, shareable document. Right after that, a separate Google Drive node uploads the PDF to the folder Aldo uses for all his audits.

The audit is proofread, formatted, exported, and filed away. Time to run the next one 🙂‍↔️

Lightweight alternative: Audit webpages using Manus

Unlike Aldo, not everyone enjoys the thrill of tinkering with n8n at 7 a.m. That is why he also showed us the alternative every marketer can use, no n8n skills required.

Good news for the non-techies out there. A tool like Manus lets you generate slide-ready audits without touching a single node. You create a knowledge bank with your brand guidelines, your TOV, and any previous audits or frameworks you want the model to follow, and it will generate a full presentation.

Compared to the n8n version, you get fewer customisation options and a higher price tag, but the process is much faster and the output comes with far better design out of the box. Use it to audit single webpages or full websites; the power is yours.

We have also included a shorter prompt you can use for Manus 🤖

🤖 Tools powering the Workflow

If you already use n8n or plan to build more automations, this workflow is a great place to start. If not, getting n8n just for this one use case is overkill.

🎢 Highs, lows and Workflow warnings

This is a shockingly simple workflow running on an overly complex platform. Most days it behaves like a well-trained assistant. On others, it reminds you that n8n is a playground where things can break just because you looked at a node the wrong way.

✅ What shines

• A repeatable, scalable, and fast homepage audit workflow

  This workflow turns a 2/3-hour homepage teardown into a structured audit that is ready in the time it takes to refill a coffee. The scoring system makes the process objective and consistent across websites, and the recommendations read like the work of a seasoned CRO consultant.

• Human expertise is injected only where it matters.

  Instead of spending hours structuring tables or copy-pasting audit templates, Aldo simply reviews the HTML output before it is converted to a PDF.

• Getting started with n8n

  If you want to break the ice with n8n, this workflow is the ideal place to begin. It is short, modular, and relatively easy to adjust. Plus, you will learn the basics of scraping, agent calls, formatting, and file export (yay!).

❌ What doesn’t shine

• n8n requires a steep learning curve

  Even though Aldo has been experimenting with hundreds of AI tools over the years, he openly admits that n8n is a hard nut to crack. “Whenever you see an error warning, it’s super hard to figure out what the problem is. Did you choose the wrong output or the wrong authentication? Debugging can drive you crazy.”

• Editing HTML output is a pain

  The workflow goes straight from HTML to PDF, which means the HTML node is where Aldo fact-checks and tweaks the agent’s output.

⚠️ Workflow warning

To solve the pain of editing output in the HTML node, we tried to vibe-code a Google Docs node on the fly during the interview (are we cool now?) using n8n’s AI assistant. It went nowhere. With n8n, the moment you start tweaking nodes, you feel the steep learning curve. Complex workflows or custom nodes often require JavaScript or Python, and while you can eventually hack your way through with ChatGPT, Claude, Gemini, or the n8n community, it still takes time to get right.

This isn’t a warning against non-technical builders. It is a reminder that behind each shiny workflow there are plenty of unsexy bits nobody mentions, like setting up API infrastructure, configuring HTTP requests, debugging your vibecoding attempts, and more.

✨ The Goldflow

“AI has freed up a lot of my time, but that’s nothing compared to how much it’s feeding my curiosity.” For Aldo, every workflow and every experiment becomes a doorway to ten more ideas he feels the urge to explore immediately. “Today I am a movie director with Sora… and tomorrow who knows? This is SO exciting for me!” You could genuinely hear the spark in his voice.

The more possibilities open up, the faster we all start tinkering with things that are fun, pointless, and suddenly: two hours. Gone. This is the paradox of our AI era.

Building workflows is only one skill. Knowing when to close the laptop is the other.

But only for now. In two weeks, you will be back here stealing the next roaring workflow we cooked up for you.

Sara Stella & Diandra ✌🏼

Can’t survive two weeks without us? Catch up on past Workflows and steal the playbooks you missed.

🚀 The LinkedIn Content Workflow for scaling thought leadership

🗣️ The Voice of Customer Workflow to power your content strategy

🤝🏼 The Buyer Validation Workflow to align Marketing and Sales

🔥 Help us roast more webpages. No violence involved.

Share

A few months ago, Search Engine Land (SEL) did something most publishers and creators avoid: they told the world they were using AI to write a section of their content…and the internet exploded.

Some accused them of killing journalism, others of secretly training models on contributors’ content. A few applauded the transparency.

Behind the storm was an editorial experiment led by Kyle Morley, our guest today. He heads Acquisitions at Semrush and Sales and Marketing at Third Door Media, the publisher of SEL and MarTech (both now part of Semrush. Yes, it’s all connected!).

Their goal was to test how far AI could go in speeding up the production of their evergreen content section without sacrificing the quality and tone their readers love.

To solve that challenge, they built Harmony, a hybrid workflow where AI drafts and humans refine, cutting production time from weeks to days.

This issue gives you a rare look at how a global brand scales content with AI and the bigger questions that come with it.

How much should you scale?

How do you keep quality intact?

How do you communicate AI usage transparently?
How should you adjust your current content ops and team?

…and many more.

Subscribe now

✅ What you’ll learn

• What an AI-assisted editorial process really looks like inside a top publication

• How humans and AI share the work across the content pipeline

• SEO writing guidelines that will help you nail that first draft with AI

🧨 What triggered The Workflow

SEL has been publishing trusted SEO and PPC insights since 2006. They cover daily industry news and analysis, trends, expert columns, and the in-depth evergreen guides marketers keep bookmarked for years.

That scope is both a blessing and a curse.
News needs constant coverage and sharp opinions, not something you hand over to AI.
Meanwhile, each long-form guide took three to four weeks to go from outline to publish, bouncing between strategists, writers, and editors. With more than 1M marketers reading every month, the team couldn’t keep up with demand.

Kyle and his team were facing a tricky crossroads.
On one hand, hiring more writers would have boosted output but risked undermining consistency in tone, structure, and quality. Each new writer meant more rounds of edits and reviews, stretching an already slow four-week cycle even further. Instead of solving the problem, scaling the writing team would have only scaled the bottleneck.

On the other hand, keeping the status quo wasn’t an option. Demand for evergreen SEO guides was growing faster than their ability to produce them.

So they turned to AI to solve the dilemma. Instead of expanding the writing pool, they re-engineered the process around editors, the subject-matter experts closest to the audience.

For this type of evergreen content, AI proved surprisingly good at research and first drafts, while editors keep full control over structure and accuracy, adding the expertise and perspective robots can’t fake. The result is the hybrid workflow we’re unpacking today: part curation, part original insight, without reinventing the wheel each time.

And here’s the part we truly love: the efficiency gains went straight back to the humans. Cutting back on external writers freed up budget to increase editors’ salaries, hire real experts, and turn great writers into even better editors.

🛠️ How to build The Workflow

Full disclosure: Harmony is a custom setup built with the help of Christopher J. DiMarco, the automation wizard who stitched together their stack and content process through n8n. This machine is complex, but there’s plenty to take away about the perfect pairing of human expertise and AI scale.

The workflow starts with entering a target keyword in a monday.com form and ends with a 5,000+ word evergreen guide ready to publish (and to rank).

We’ll also share the SEO Content Writing Guidelines that Kyle’s team baked into the system, so you can tweak them and run.

Let’s break it down.

Prep: Setup and guardrails

Building a workflow in n8n is the easy part. The hard part is having a well-oiled content engine worth scaling.

Kyle and his team at SEL had 15+ years of high-quality content to lean on, but the AI was trained only on the new guide content format launched in late 2024, paired with detailed editorial guidelines built exclusively for that purpose.

They also drew a hard line: not all content should be touched by AI. Harmony was designed only for evergreen, educational, SEO-focused guides with a fixed structure and strong research base, while news and thought leadership remain fully written by humans.

Each guide still touches five people: a content strategist, an editor, a senior editor, a project manager, and a designer. The bottleneck wasn’t the writing, it was the waiting. Drafts bounced back and forth for weeks, collecting dust between revisions.

With such a clear process in place, this was the perfect use case to let AI do more of the good work.

Step 1. Keyword research

The magic starts with one keyword.

The content strategist drops it into a simple monday.com form, which triggers automated research. Using Perplexity and Semrush data, Harmony analyzes input keywords to identify topic category, search intent, target audience, and related keywords, then logs everything in a Google Sheet for tracking.

Behind the scenes, Harmony also checks for internal links across SEL and MarTech to avoid content cannibalization. Once the research gets a human thumbs-up, the automation creates a monday.com task and moves the piece to the next stage: creating a killer outline.

Step 2. Outline creation

The outline generated by the system includes section headings, subheadings, and notes on what each part should cover. The goal is to give editors and strategists a clear blueprint before any writing begins.

The outline is stored in Google Docs and automatically linked to monday.com, where the task status updates to “Outline ready.” From there, the editor reviews, tweaks, or requests changes before marking it as approved, which automatically moves it to the next stage: writing the first draft.

Step 3: Let’s write! 📝

Once the outline is approved, Harmony hands things over to ChatGPT, which turns it into a full draft. Each section expands into clear, structured paragraphs, automatically formatted in Google Docs and linked back to monday.com for review. Every status change sparks the next AI workflow, completing the piece without anyone (but AI) lifting a finger.

But here’s where the real magic happens. The AI writes using SEL’s Content Writing Guidelines, following the same principles as their editors: chunkable paragraphs, set structure, and a conversational “coffee with an SEO pro” tone.

Editors then jump in to sharpen facts, inject their expertise, and ensure it sounds…human.

If a section needs more work, they simply flip the status to “Rewrite”, and Harmony runs another AI pass with the new edits.

😎 Grab the exact writing guidelines Kyle’s team uses to create high-quality blog drafts with AI

Step 4: Final approval & go live

When the editor marks a draft as “Approved” in monday.com, Harmony automatically moves it to the final review stage. The senior editor does one last pass for accuracy, consistency, and formatting before it goes live.

Finally, the system pushes the draft to WordPress as a preview post and updates the monday.com board. From there, hitting publish ismanual by choice. It’s the team’s final human checkpoint before content reaches SEL’s 1M+ readers.

🤖 Tools powering the Workflow

Despite its complexity, this workflow is built on a simple idea: let every tool and every human do what they do best.

Everything runs through n8n on the back end and monday.com on the front end, where each status change automatically triggers the next AI workflow. Research, outline, draft, and rewrite move seamlessly without manual handoffs. Editors keep full control of the process, stepping in to course-correct the AI’s work when needed.

🎢 Highs, lows and Workflow warnings

In a workflow where humans and AI finally coexist peacefully, something is still bound to go wrong. Harmony sounds…ehm…runs beautifully most days, but it still hits a few sour notes here and there.

✅ What shines

• Human control at the speed of AI

  Production time dropped from 4 weeks to about 1 week without sacrificing content quality and consistency. “It still takes several hours of intense editing, but an editor could produce search-optimized 5,000+-word guides in a single day if they wanted to.”, Kyle told us.

• Doubling down on human expertise

  Another important side effect of the workflow is how it reshaped the team composition. Many former writers now work as editors, refining AI drafts instead of starting from scratch. As AI gets better at first-draft writing, human judgment, expertise, and tone matter even more. “We’re saving money overall, but now editors on our team are better compensated than they ever were before. The pie is now distributed differently.”

• Sweet spot between automation and human oversight

  AI handles research, outlines, and first drafts. Editors review, refine, and add expertise where it matters most. Every step still requires a human greenlight, ensuring quality never falls out of the loop.

❌ What doesn’t shine

• Complex setup and maintenance

  It takes solid technical know-how to build and maintain Harmony, and one small error in n8n can disrupt the entire editorial pipeline. This is an advanced system built for companies operating at publisher-level content velocity.

• Not for every content type

  This workflow shines when the rules are clear and (most) facts already exist. It’s perfect for SEO guides and educational pieces built on solid research, but not for news or thought leadership that rely on originality, timing, and fresh ideas.

• AI ethics & other minefields

  When SEL revealed it was using AI to write part of its content, the conversation heated up fast. Some readers assumed the models were trained on contributor work without consent. It wasn’t true, but it proved how sensitive AI ethics are and how crucial transparent, precise communication really is.

⚠️ Workflow warning

Harmony helped the team reach new heights in content marketing. The next step is not to stop, it is to calibrate. “We’ve covered the majority of the topics we wanted to cover. We’ll slow the pace a bit soon.” Finding the right scale is just as important as finding the right workflow.

✨ The Goldflow

“This project reaffirmed the importance of humans in the loop and made us feel less scared about what AI is doing to our industry. We gave it so much direction, so many steps to get to that first draft, and it still isn’t good enough until people touch it.”

AI may write the words, but humans give them meaning. Even advanced workflows like Harmony aren’t built to replace people, but to prove that the future of marketing is built when humans and AI work together… In harmony.

What’s your plan for Thursday in two weeks? Right, coming back here to steal our next workflow.

Sara Stella & Diandra ✌🏼

✍🏼 Let’s remind the world that AI can write content, but only humans make it worth reading.

Share

🌯 It’s a major wrap

This was the ninth edition of The Workflow, a bi-weekly newsletter by Sara Lattanzio & Diandra Escobar, where we unpack the playbooks that actually move the needle for marketers smarter than their tools.

Subscribe now

This week’s guest, Sarah Adam, Head of Influencer Marketing at Wix, spends her days building long-term partnerships with hundreds of creators across every social platform you can imagine. Which is why the rise of AI influencers and “AI twins” freaks her out (same for us, by the way). “It’s hard to relate to someone who’s made of pixels and looks like a model,” she told us. “Influence only works when it feels human.”

The workflow Sarah shares today is her quiet rebellion in the name of authenticity. It uses AI to find the right creators faster, not to replace them.

If influencer marketing has a place in your 2026 strategy, you’ll want to steal this one.

Subscribe now

✅ What you’ll learn

• How to use AI to speed up influencer discovery without killing authenticity

• How to extract keywords for influencer discovery from product documentation

• How to define ideal influencer profiles for several products and different ICPs

🧨 What triggered The Workflow

There’s no such thing as a one-size-fits-all influencer marketing approach at Wix. With several brands in their portfolio targeting very different user groups, each campaign needs its own formula. Some speak to professional designers building sites at scale with Wix Studio, others to individuals or small businesses building apps with Base 44, Wix’s recently acquired AI app builder for non-techies. The first group looks for depth, tutorials, and design inspiration. The second values snackable content, simplicity, and creators who show that tech can be accessible to anyone.

TLDR; Each audience needs its own story, tone, and influencer type to make the product click.

The prerequisite for winning influencer campaigns is crystal-clear product marketing. Sarah starts with a detailed deck outlining features, positioning, and target users. From there, she has three main jobs: defining influencer profiles that each audience will resonate with, finding those influencers, and making the deals happen.

To identify the right cleators, Sarah starts by clustering keywords related to the product and the audience’s topics of interest.

Before she created the Influencer discovery workflow, she had to chase the SEO team for help (spoiler: not their top priority) or spend hours buried in decks and demos, manually extracting keywords and hoping they might unlock the right creators.

Today, she just drops the product marketing deck into ChatGPT and instantly builds hyper-relevant influencer profiles based on topical keyword clusters.

🤯

🛠️ How to build The Workflow

It all starts with the product marketing deck, the single source of truth that defines features, benefits, and ICPs. From there, the workflow runs through three simple prompts: one to extract keywords, one to map them to ICPs, and one to reverse-engineer influencer profiles for each audience.

Step 1: Gather solid product and brand documentation

This step decides how good your output will be, so don’t half-bake it. For Sarah, it starts with a 20-slide deck that spells out what the product is, who it’s for, and why anyone should care.

Strong product marketing documentation includes:

• Product overview – what the product or feature is, what problem it solves, and why it matters. Includes taglines, one-liners, and a short narrative that explains the product in plain English.

• Core pillars and features – the product’s main values or themes, mapped to the features that deliver them, plus short descriptions of what each feature does and the value it brings.

• Benefits and USPs – the real outcomes for users, such as efficiency, productivity, or creativity.

• Market context – competitive landscape, positioning statements, and key talking points often found in the press release.

• FAQs and objections – common questions, hesitations, and how the product addresses them.

• ICP profiling – who you’re targeting: roles, goals, pain points, and where they spend their time online.

That’s your product from A to Z, ready for ChatGPT ingestion.

👉🏼 Grab all the prompts you need to run this workflow before you end up keyword-hunting at midnight.

Step 2. Turn product documentation into relevant keywords

Time to turn that product deck into insights. Upload the documentation from Step 1 into ChatGPT and run the first prompt to extract and expand a list of keywords that describe the product, its features, and the topics your audience actually cares about.

According to the prompt, each output keyword will be short (a maximum of three words) and grouped by theme. These clusters will become the building blocks for your influencer search later.

Step 3: Map keywords to different ICPs

Now that you’ve got your keyword clusters, it’s time to give them direction. Use the second prompt to define your Ideal Customer Profiles (ICPs) based on the brand documentation and assign the right keyword groups to each one.

⚠️This step is especially powerful if you manage multiple products and ICPs, as it gives every campaign its own audience playbook and shared vocabulary.

Step 4: Create influencer profiles per ICP

With your ICPs defined and keyword clusters ready, it’s time to bring the humans back into the mix. Run the third prompt to generate ideal influencer profiles for each ICP. ChatGPT will combine the audience insights and keyword themes to describe the types of creators each buyer group actually trusts.

You’ll get scarily detailed influencer blueprints covering platform mix, content style, audience size, and tone. For example, one profile might describe a YouTube designer creating tutorial-style content, while another might highlight a TikTok creator focused on AI app hacks for beginners. Sometimes, ChatGPT will even estimate (fictional) posting cadence or content performance data for specific creator profiles.

🤳🏻 Now all you have to do is go find those creatures in their natural habitat: social media.

Step 5: Find and vet influencers

This is where AI hands the baton back to you. Use the keyword clusters from Step 2 to search for creators in your discovery tools or directly on social platforms. Sarah uses the influencer discovery platform Favikon, but you can also search manually. The difference is granularity (filters), time, and access to audience or performance insights.

Once you’ve pulled a list of potential matches, switch to the influencer ChatGPT-generated blueprints from Step 4 to validate them.

Ask yourself the following questions when evaluating influencers:

• Do they match the tone, platform mix, and content style your ICP actually trusts?

• Are they credible and consistent in what they post?

• Would your audience actually listen to them?

• Is their content and tone aligned with your brand?

• Does their engagement look authentic?

The goal isn’t to find superstars with billions of followers. It’s to find people your ICP already trusts.

Finally, reach out and make the deal happen. AI gets you here, and the human touch takes you further.

🤖 Tools powering the Workflow

Sarah’s stack is the good kind of boring: zero automations, just a setup that gets the job done so she can focus on the tasks AI can’t replace. What truly matters to her is building long-lasting partnerships with creators who genuinely align with the brand and keep its story alive, campaign after campaign and launch after launch.

🎢 Highs, lows and Workflow warnings

Forget human-assisted AI. This is AI-assisted humans at their best. Sarah’s workflow is so dialed in that there’s barely anything to fix — maybe just a solid base to build on.

✅ What shines

• Time halved, precision doubled

  The workflow cuts out two to four hours of manual keyword hunting per campaign. “Now I just refine what AI gives me instead of spending an afternoon digging through product decks,” Sarah said.

  It’s not just about time savings, though. “If I did it manually, I’d never be able to build such a comprehensive list of keywords,” she added. Even if she doesn’t use every keyword to hunt for influencers, the breadth gives her room to experiment, find fresh creators, and launch campaigns faster across product lines.

• Clean handoff between human and AI

  The workflow stops exactly where it should. AI lays the foundation, but Sarah and her team take over when real people enter the picture. From outreach and negotiation to briefing and approvals, she keeps every interaction personal. “I know it can be automated, but relationships with influencers are too important to us.” That’s the line she won’t cross.

❌ What doesn’t shine

• Manual influencer review eats up tons of time

  Even with AI handling the heavy lifting, Sarah still has to validate every creator profile manually. “I can’t just export influencer lists and fire emails. I have to go one by one, open each profile, and check their posts to see if they’re actually relevant.” Favikon’s filters help a lot, but sometimes creators show up in categories that aren’t their core focus. This is a slow, repetitive step she’d gladly hand over to AI, but it’s necessary to ensure every partnership truly fits.

⚠️ Workflow warning

Resist the hype to scale and automate everything. Just because you can, does it mean you should? When we asked Sarah if she’d ever considered automating parts of this workflow, like importing or exporting data from ChatGPT, her answer was simple: nah.

“I use it when I’m working on something brand new, which doesn’t happen every day. The research we do now will probably run for at least six months.” Hard to argue with that.

Still, a custom GPT would probably save her a couple minutes of prompt copy-pasting per quarter at the very least. 🙃

✨ The Goldflow

Working with AI reminds Sarah about what’s truly valuable in her day-to-day. The hours she spends messaging creators, negotiating, reviewing drafts, and giving feedback aren’t inefficiencies; they’re THE job. What AI took over were the prerequisites.

It’s the same tension every marketer faces today: deciding when to delegate to machines and when to protect the human touch intentionally. As Sarah said, half-joking (but not really), “If one day I email an influencer and an AI agent replies, that’s when I’ll quit and become a kindergarten teacher.”

…Assuming, of course, we’ll still be able to tell it’s an AI agent.

See you in two weeks,
Sara Stella & Diandra ✌🏼

🥊 Help us stop the rise of AI influencers with perfect teeth.

Share

🌯 It’s a major wrap

This was The Workflow, a bi-weekly newsletter by Sara Stella Lattanzio & Diandra Escobar, featuring one real workflow that moved the needle for someone even smarter than the system they built.

Subscribe now

From creating a marketing report to generating blog article briefs, we’ve shared quite a few content workflows, but never one that pokes holes in your content before the CFO does.

Andrei Zinkevich, co-founder of ABM consultancy Fullfunnel.io, has seen 50 shades of sales and marketing misalignment throughout his career. The usual story: Marketing cranks out content on a streak, and Sales won’t touch a single piece (sometimes they even pretend it doesn’t exist).

The reason is that most content for high-ACV businesses is disconnected from real customer pains, recycled from the top ten Google results, and created in isolation without input from subject matter experts. Marketing is too far removed from customers and AI has only made the problem worse. “No sales rep will ever send anything that makes them look like an amateur to a prospect. As simple as that.”

To stop the bleeding, Andrei built what he calls a synthetic buyer panel, an AI-powered feedback loop that reviews and challenges your content from the perspective of your actual target buyer.

His workflow mirrors his ABM approach: define ICPs based on closed-won deals, segment buying committees into personas, and create content that speaks to buyers’ pain points and supports long sales cycles. Not another Frankenstein PDF bloated with gut-feeling edits and marketing fluff, you heard that right.

🥜 Let’s crack it open.

Subscribe now

✅ What you’ll learn

• How to review your content through your buyer personas’ eyes using AI

• How to make every asset reflect your ICP’s real pain points and priorities

• How to rebuild Sales–Marketing alignment with content that actually moves deals

🧨 What triggered The Workflow

Most B2B marketing teams create content to please algorithms instead of helping buyers, or hide whitepapers behind forms to chase MQLs. This outdated playbook became the industry standard, with few marketers actually measuring whether their content drives pipeline or helps close deals faster.

Despite popular belief, adding more AI tools on top only makes the problem worse. The real solution is a reality check from the market itself. Andrei’s workflow uses AI to bring the buyer’s voice back into the process, not bury it deeper in keyword soup.

👇🏼 If you think it sounds too good to be true, wait till you see the step-by-step.

🛠️ How to build The Workflow

Andrei’s Buyer Validation Workflow is the grown-up version of “let’s see what the market thinks.”

The panelists are AI-simulated buyers trained on real ICP data. The setup couldn’t be simpler: one Claude project, no automations (and zero shame in admitting that).

What follows is the playbook (prompts included!) you can use to build a validation engine around your own buyer personas.

Step 1: Build your knowledge base

This might be the simplest workflow we’ve ever featured, and that’s exactly why it works.

The beauty of AI is that you can train it to act like your customer…if you actually know them.

Andrei’s knowledge base is built from real, customer-sourced inputs (in his case, B2B marketing leaders) that reflect how buyers think, talk, and make decisions.

Here’s what goes in:

• Buyer profiles from your top five fastest and five biggest closed-won deals. Each includes the champion’s name, title, and LinkedIn biography, copied verbatim or saved as screenshots.

• Strategic challenges gathered from client surveys and discovery calls. Answers to questions like “What’s really blocking pipeline right now?”

• Voice-of-customer quotes lifted straight from transcripts. No paraphrasing, no cleanup.

• Anecdotes that describe real situations, not hypotheticals. He loves lines like “Our regional marketers inflated MQL numbers just to hit the dashboard.” They make the AI’s feedback painfully specific.

When we asked if he automates the data extraction to create these assets, Andrei just laughed out loud:

“It’s fancy to talk about automations, but often maintaining them is way more complex than doing the work manually.”

At FullFunnel, his project manager gets a notification when a transcript is ready, downloads it, and pastes it into the dedicated Claude project. It takes 30 seconds. The same person then shares Claude’s output in a common workspace so everyone can see the insights.

“What we want is knowledge transfer,” Andrei said. “The more people get access to it, the better. In big companies, most marketers have never listened to a single customer call. What automation will fix that?”

The sharper the knowledge base, the smarter your “buyers” will sound when they start reviewing your content.

👉🏼 Grab the Transcript filtering prompt to build a Claude project (or Custom GPT) that extracts client insights for your LLM knowledge base, then move to the next step.

Step 2. Set up a Claude project for your synthetic buyer panel

With your knowledge base ready, it’s time to give it a home. Start by creating a new Claude project dedicated to a specific member of the buying committee (or focused on a particular product use case).

Here’s how to set it up:

1️⃣ Create the project: In Claude, start a new project and name it after the buyer persona or use case you’re targeting (e.g., “B2B CMO panel” or “Cold outbound panel”).

2️⃣ Add your knowledge base: Drop in the full set of documents we mentioned in Step 1.

3️⃣ Add persona-specific instructions: They make Claude act like a real buyer who doesn’t let lazy claims slide. They should include your buyer’s background, daily pressures, decision-making criteria, and the exact questions they’d ask when evaluating your content.

👉🏼Grab the Buyer persona prompt and use it as a blueprint for your Claude project instructions.

Step 3: Upload your content and brace for feedback

Drop your blog, case study, or campaign asset into Claude and let the roasting begin.

Your AI-simulated buyer will scan it for fluff, claims that are detached from reality, and missing business logic. The feedback might sting, but that is exactly why this gem exists.

If your synthetic reader tears it apart today, your real buyer will not tomorrow. And when Sales finally shares it, it will be an asset that gets replies instead of awkward silence.

🤖 Tools powering the Workflow

Andrei keeps his stack refreshingly simple. Google Meet records and transcribes customer calls that feed the knowledge base, then Claude runs the synthetic buyer panel. The final and most important tool is still your brain, because you need to know your customer, craft the right prompts, interpret the feedback, and decide what truly deserves a rewrite.

🎢 Highs, lows and Workflow warnings

Simplicity wins, but even the best systems need tuning. Here’s where this week’s workflow shines and where it still creaks.

✅ What shines

Fast validation without the noise
Andrei built this workflow to cut through “Frankenstein content,” assets stitched together by committees and stripped of any real buyer perspective. “Everyone adds their two cents until the message is dead.” Now a single Claude review gives anyone, even the know-it-alls, a quick reality check grounded in buyer logic rather than internal politics.

Shared context, sharper messaging
Centralizing customer insights and AI-simulated feedback can transform how teams collaborate. “Before, marketing had ideas, sales had opinions, and nobody listened to customers,” Andrei said. “This workflow lets everyone start from the same customer knowledge.” The result is content that sounds like it came from people who talk to buyers, not just talk about them.

Fast track to real insights
Three years ago, validating content meant recording calls in Google Meet, uploading them to Descript, and manually reviewing transcripts to find quotes and pain points. Market feedback came later through LinkedIn reactions, comments, and webinars, which remain valuable signals from the audience.

What changed is the speed and precision before publishing. With the synthetic buyer panel, Andrei can test ideas upfront and ensure they align with what buyers actually care about. “It’s not automation for the sake of it’s acceleration of what already works.”

The time to validation, meaning the moment when a message truly sticks with the buyer, has dropped by about 90 percent.

❌ What doesn’t shine

AI makes bad inputs even worse
If your transcripts are shallow or your personas are made up, the output will be just as hollow. “AI is an accelerator. If your process sucks, it will just accelerate…the suck.”

Overconfidence kills learning
It’s tempting to take Claude’s feedback as gospel, but Andrei warns that it should start a conversation, not end one. “The idea is to make people think, not to replace their thinking.”

⚠️ Workflow warning

Too many teams still work in silos, each chasing a different version of the customer. The ultimate goal is to create a single, validated buyer profile that everyone agrees on and uses consistently across marketing, sales, and leadership. Without that shared foundation, even the best workflow will collapse.

✨ The Goldflow

The simplicity of this workflow holds the real lesson. The power isn’t in building 40 AI agents to run your marketing. It’s in clear processes, precise ICP profiling, and a shared knowledge bank built manually from existing deals and real conversations.

Run this workflow with a mediocre customer profile and it becomes useless, extra work that adds a layer of AI slop on top of already bad inputs.

Two weeks from now: another workflow, another reminder that speed only matters when you’re pointed in the right direction.

Sara Stella & Diandra ✌🏼

🥊 Help us fight the epidemic of fluffy content!

Share

The Workflow is free to read, but requires a lot of work to assemble. Subscribe to support our work so we can keep the AI noise low and the insights high.

Subscribe now

In fact, this issue wasn’t even planned.

But starting the year without saying thank you didn’t feel right.

By the end of 2025, 1100 of you trusted us with your email addresses. Inbox trust is sacred real estate these days. We couldn’t just… disappear into January pretending that didn’t happen.

So this is us. Two humans behind keyboards, writing to say: THANK YOU.

How the Workflow started workflowing

The Workflow started as a survival mechanism to cope with the AI hype and somehow turned into a newsletter with 1100 marketers, growth leads, founders, and CMOs across the US and EU loving it. People we truly admire ask us to be interviewed and brands reach out to ask if we’re accepting sponsorships

…which is wild at best with just 6 issues live (!)

The first couple of interviews helped us refine the newsletter format. That’s where having friends with deep AI knowledge (and a lot of patience!) really helped.

Just like people who build products to solve pains they once experienced, we literally built our own space to learn AI from and with our community, because everything else felt like a race to stack more nodes in n8n for no particular reason.

Everywhere we looked, AI was being presented as either magic or menace. Endless tool lists. Loud promises. Perfect outputs, with zero context for the work behind them. What was missing were the receipts, the “but what’s the goal?”s, and the messy middle between ideas and results.

We wanted to document that messy middle to show how marketers and operators (like you and us) use AI and automation when deadlines, stakeholders, and imperfect data are part of the game.

This attitude shapes (and will continue to shape!) everything we publish, for as long as this newsletter exists.

🌍 Also: geography did not stop this

One more thing we love about this project.

Diandra lives in Colombia.
Sara Stella lives in Switzerland.

Different continents, same allergy to AI hype, LinkedIn gurus, and marketing fluff. We’ve grown closer than many people who meet daily, simply because we share the same worldview and…hate the same things, probably?

One of the highlights this year was finally meeting in person at Semrush Spotlight and realizing three amazing things:

✅ Yes, we both have legs (phew)
✅ Yes, we both exist (double phew)
✅ Yes, the other one is even better in person (triple love)

What we’ve workflowed so far

In 2025, we stuck to our bi-weekly shipping cadence and published 6 issues.
Okay, there was also a pilot Manifesto issue that stated our intentions with this newsletter, but that doesn’t count.

From simple prompt chains to workflows that made us rethink what’s possible, we’ve debunked the myth that AI needs to be overly complicated to be useful.

If you want a head start on your New Year’s resolutions, this is the perfect moment to catch up on the issues already live and maybe start building something small when you need an excuse to avoid your relatives.

📚 The Marketing Report Workflow by Sara Stella Lattanzio

🧮 The Lead Scoring Workflow by Vukasin Vukosavljevic

🏅 The Brand Positioning Workflow by Clay Ostrom

📝 The BOFU Article Workflow by Lashay Lewis

📞 The Call-to-LinkedIn Content Workflow by Grant Hushek

🗣️ The Voice of Customer Workflow by Mina Mesbahi

Expect many more non-AI gurus next year. How many of those do you recognize?

❓What type of workflow should we cover next?

🎁 The important stuff: your gift

Only ~20% of readers actually click on newsletter links. That’s a shame, considering we share templates or prompts in every single issue of The Workflow.

These prompts are essential if you want to replicate one of our workflows, but they’re just as useful if you want to solve one specific task with your LLM of choice (ChatGPT, Perplexity, Claude, Gemini… we don’t discriminate).

Ready to unwrap the paper? 3,2,1…

Enter the Prompt Vault: the best prompts from all our issues so far.

• 15 prompts

• Ready to Ctrl + C / Ctrl + V straight into your LLM of choice
  

→ Get the Prompt Vault Here! 🎁

👩🏼‍🍳 What’s cooking

We’re currently on a small hiatus (which we happily broke with this issue).
Read: Sara’s stuffing her face with excellent Italian food and Nano Banana-ing family pics, while Diandra is forcing her whole family to take 20 supplements a day and explaining the ROI of magnesium glycinate at dinner.

The first 2026 issue of The Workflow drops on Jan 8th, where we’ll tackle one of the hardest problems out there: simplifying Sales and Marketing alignment.

Enjoy the break. AI still can’t do it for you.

Thanks for your early support. It means a LOT. ❤️

Sara Stella & Diandra ✌🏼

🙏🏼 P.S. If our newsletter was useful to you this year, consider forwarding this to someone who’d appreciate fewer manual tasks in 2026.

Share

In our last issue, our guest Grant combined multiple LLMs, automation, and human review to transform his endless Zoom calls into a daily pipeline of LinkedIn posts that sound exactly like him.

This week, we’re flipping the script. Same raw material, transcripts, but a completely different use case.

Sara Stella’s friend Mina Mesbahi, founder of boutique content consultancy Digichapter, built a CustomGPT (baptism name: Customer Echo 🫨) that digs through sales call transcripts (discovery, pitches, demos etc.) with Swiss precision, extracting customer insights to fuel a winning content strategy.

Time to trade clickbait hacks for what your buyers actually care about. You ready?

Subscribe now

✅ What you’ll learn

• How to build and train a CustomGPT to extract customer pains, goals, and language directly from sales call transcripts

• How to automatically organize emerging insights in a centralized knowledge hub for better analysis

• How to ground your content strategy in what buyers actually say

🧨 What triggered The Workflow

For B2B marketers, pitches, demos, and customer call transcripts are goldmines that mostly go untouched. The old problem was access: marketing rarely sat in on sales calls, and relying on someone else to share insights was a losing bet. Today, the problem is the opposite: endless recordings with no system to turn them into repeatable, actionable insight.

Mina spotted this gap again and again: “Teams have a ton of insights in their sales conversations, but it’s either not being operationalized or they just don’t have access to it.”

Her remedy was to build a CustomGPT that pipes transcript insights into Airtable, surfacing customer pains, goals, and phrasing at scale.

🌟If you’re already sitting on this kind of gold, how long can you afford to leave it buried?

🛠️ How to build The Workflow

Mina’s Customer Echo works like a sharp-eyed virtual analyst, combing through transcripts and tagging every quote by pain, goal, trigger, or objection. Once that engine is in place, the rest clicks together: capture the calls with Fathom, run transcripts through the GPT, drop structured quotes into Airtable, and watch themes roll up into insights that actually steer your content strategy. What follows is the playbook, complete with GPT instructions and Airtable views — so you can mirror the system, giving your transcripts a second life.

Step 1: Build your CustomGPT

Mina nicknamed her CustomGPT Customer Echo because, while it can handle many tasks, its only real job is to mirror the customer’s voice back to the team.

Set up instructions for your own Customer Echo

1. Knowledge bank

When configuring your GPT, load it with a knowledge bank that anchors it in your strategy. Upload a content strategy foundation doc containing:

• ICP details (demographics, firmographics, pains, jobs-to-be-done)

• Content goals (translated from business prios)

• Company messaging and perspectives

• Content themes

• Content inventory (past + planned) as a blueprint

• Promotion and campaign plans

• Any other strategic sales and marketing inputs

2. Recommended model

Mina prefers GPT-4o over 5 because it’s faster, handles conversational data more smoothly, and captures nuance in how people actually speak — quirks, hesitations, even the emotional weight behind objections. Perfect for parsing transcripts where customer pains are often hidden between the lines.

3. Capabilities

Leave web search off. You want the model laser-focused on your transcripts and strategy docs, not wandering the internet.

4. Instructions and conversation starters

These live in a separate doc. They guide the analysis and dictate how outputs are structured. Think prompts like comparing quotes against strategy topics, gap analyses, or surfacing ready-to-use VoC examples. Mina prefers to handle formatting rules inside the CustomGPT itself and as you’ll see in the next step, she slices insights into four fixed Airtable views.

👉🏼 Grab the Conversation starters & Instructions doc to build your CustomGPT

Step 2. Capture the calls

Once your CustomGPT is ready, you need a steady stream of transcripts to feed it. Mina uses Fathom to record and transcribe sales or customer calls. From there, the raw transcript (timestamps and all) flows into Airtable via Zapier. If you’ve never heard of Aitable, drop it into Notion, ClickUp, or even Google Sheets.

The only rule: pick one home base your team can actually use as a single source of truth.

Her advice: don’t sanitize transcripts too early. Keep timestamps intact so you can always trace a quote back to the exact moment in a call. If privacy is a concern, you can strip names, but avoid cutting context that could matter later.

How the flow works

1. Customer call ends → Fathom transcript ready.

2. Zapier pushes transcript + metadata into Airtable.

3. Airtable creates a record with the raw transcript attached.
   That record then triggers your CustomGPT to process it and return structured quotes.

Step 3. Run the CustomGPT and extract the gold

When a transcript is ready, Zapier sends it straight to your CustomGPT. This is where the real analysis begins, and it’s not one-size-fits-all. The output depends on the conversation starter

In this case, we’ll switch between three:

• Alignment check: “Compare these customer quotes with my content strategy topics and highlight what’s aligned vs. not covered.”

• Gap analysis: “What customer needs from this call don’t map to our current strategy?”

• Voice of customer examples: “Suggest three relevant verbatim quotes we could use directly.”

Each run gives you a different lens on the same transcript. The goal at this stage is to pull out structured insights you can work with later.

Once the CustomGPT has done its job, Zapier sends the generated data back into Airtable. That’s where the real sorting and pattern-spotting happens — which we’ll break down in the next step.

Step 4. Analyze the data in your central knowledge hub

With transcripts processed by your CustomGPT, Airtable (or your PM tool of choice) becomes your control room. The workflow is structured into four tables, and within each table you can spin up different views using filters and groups to apply the lens you need.

1. Voice of the customer (VOC) view

Conversation starter: “Suggest 3 relevant voice of customer examples.”

This is the raw library of tagged quotes that highlight pains, goals, and objections in the exact words customers use. Think of it as your searchable archive. Writers and marketers can dip in any time to pull authentic phrasing for copy, headlines, or pitch decks.

2. Alignment assessment view

Conversation starter: “Compare these customer quotes with my content strategy topics and highlight what’s aligned vs. not covered.”

Here, insights are compared against your existing content strategy topics. Each quote is labeled as aligned, partially aligned, or not covered. This shows whether your current strategy actually reflects customer reality. For (content) marketers, it’s the quickest way to spot gaps before they snowball.

3. Gap analysis view

Conversation starter: “Give me a gap analysis from this call: what customer needs don’t map to our current strategy?”

This view highlights customer needs that don’t map to your current strategy at all. Marketers can use it to propose new content themes or angles. Product marketers can use this view to spot missing tutorials, customer onboarding docs, or areas where product messaging needs a refresh.

4. Sales enablement view

This table pulls data straight from the Voice of Customer library, capturing every quote the CustomGPT tags as a pain, goal, trigger, or objection, and matching them with topics already covered in your content strategy and assets.

From there, you can create a clean view tailored for sales by grouping insights by content themes, quarters, or content formats, depending on how your reps think and work. Done right, this becomes a plug-and-play library —a single place for proof points, objection-handling materials, case studies, thought leadership, and authentic customer phrases that they can use directly in calls and follow-ups.

No more excuses for not using the content you’ve created or for slacking you at 5 a.m. with “Where’s the case study??? #pitchTomorrow.”

⚠️ Mina recommends not to simply copy and paste her setup without giving it some thought.

“Within your company and team, you’re going to have different dynamics, nuances, and challenges. The power is in customizing the views and insights to fit your reality. It’s an extremely versatile workflow.”

👉🏼 Haven’t built your Custom GPT yet? Grab the Conversation starters & Instructions doc to do it right now.

🤖 Tools powering the Workflow

The Voice of Customer Workflow runs on a lean stack: Fathom for call transcripts, Zapier to move data between tools, a CustomGPT for tagging and analyzing quotes, and Airtable as the central hub for storing information. It’s a setup any scrappy team can afford. The real power lies in the custom views that keep strategy and sales anchored in the customer’s voice, as well as the ability to refresh customer insights in real-time.

🎢 Highs, lows and Workflow warnings

This workflow is gold but AI still has its cracks. Here’s where it shines and where it stumbles.

✅ What shines

• HUGE time savings: Manually reviewing a single transcript can easily take four hours: rewatching the call, combing through notes, and cross-checking your content inventory. With this workflow, the grind drops to minutes and human error to zero.

• Holistic customer view: Both teams work from a single, living library of customer insights that updates with every new call, ensuring continuous alignment.

❌ What doesn’t shine

• Short calls with little substance don’t magically become gold. Sometimes you just won’t get enough insight to work with.

• On the flip side, long or insight-dense calls can cause analysis paralysis. Don’t chase every single nugget; put your strategist’s hat on and filter for what matters this month or for X campaign.

⚠️ Workflow warning

The Voice of Customer Workflow isn’t a substitute for being in the room. Interacting with real humans provides the context and understanding you can’t get from Airtable rows alone. As Mina puts it, “Don’t let the system become a crutch. Use it to lighten the load, not replace human connection with customers.”

✨ The Goldflow

“AI won’t save you from bad habits, it will just scale them.” Mina’s reminder is sharp.

The magic isn’t in the model; it’s in the documentation. A well-structured content strategy turns AI into a force multiplier; a messy Google Drive turns it into a random quote machine. Get specific, get documented, and your GPT becomes a strategist. Skip that step, and you’re just asking the internet to make stuff up.

Two weeks from now: another workflow, another excuse to rethink how you work.
Sara Stella & Diandra ✌🏼

🫨 Don’t keep Customer Echo a secret

Share

🌯 It’s a major wrap

This was the sixth edition of The Workflow, a bi-weekly newsletter by Sara Stella Lattanzio & Diandra Escobar, featuring one real workflow that moved the needle for someone even smarter than the system they built.

Subscribe now

Last time, we cracked the code on writing high-converting product-led SEO articles with the ultimate BOFU workflow.

This week, we’re shifting gears across the funnel (and across platforms) to the one channel that’s non-negotiable in B2B: LinkedIn.

If you’ve ever thought, “I should post that,” during a client call and then never did, this one’s for you. Whether you’re a founder juggling back-to-back Zooms, a marketer ghostwriting for execs, or anyone sitting on piles of unused transcripts, we’ve got a system. Calls become spicy takes, spicy takes become drafts, and drafts land neatly in your calendar.

Posting daily on LinkedIn suddenly stops feeling like your second full-time job.

Subscribe now

✅ What you’ll learn

• How to automate LinkedIn content creation from Zoom call transcripts

• How to extract and filter post-worthy topics with ChatGPT

• How to turn founder insights into spicy, opinionated LinkedIn takes with Claude

• How to build a human-in-the-loop workflow in Slack for authentic content

🧨 What triggered The Workflow

Creating consistent LinkedIn content is where most founders stall. Grant Hushek, founder of AI and automation agency Grantbot, was no exception. He lives on Zoom, not in Docs, and the blank page kept winning. “I really needed a draft. I HATE staring at a blank page and creating content from day to day,” he told us. RIP consistent posting, even though the raw material to power his LinkedIn presence was trapped in his many calls.

So, he built Call-to-Content to handle a few repetitive jobs: scan every Fathom transcript, pull out the quotable bits, and hand him ‘spicy takes’ to approve in Slack. It’s built for founder-led LinkedIn content, but works just as well for any exec, consultant, or team sitting on a mountain of transcripts. Sounds like AI scale and efficiency, minus the slop.

👇🏼 If you think it sounds too good to be true, wait till you see the step-by-step.

🛠️ How to build The Workflow

We’re starting to notice a clear pattern: the best workflows come from solving a personal pain first, then scaling, never from duct-taping 50 AI tools together. Grant’s calls were overflowing with content-worthy insights, but none of it was making it to LinkedIn. “I built this to keep up my LinkedIn game, and then founder friends started asking for it. That’s when I realized others might even pay for it,” he told us.

That’s how his four-step workflow was born: capture & filter, humanize, generate content, schedule.

🎁 Steps 1-3 come with a prompt baked directly into his process (we’ve collected them here.)

As we break down the workflow, we’ll refer back to it so you can see exactly how Grant turns raw transcripts into ready-to-post content in minutes.

Step 1: Capture insights from transcripts

Step one is simple:

1. Have more meetings than you can humanly handle.

2. Catch every insight from sales calls, client syncs, and internal stand-ups.

Grant uses Fathom to auto-record and transcribe Zoom calls. From there, Zapier takes over, piping the transcript directly into ChatGPT and running two prompts in sequence:

1. Topic identification — ChatGPT scans the transcript for unique insights and answers Grant gave in calls that could be perfect post material. It outputs each as a list separated by @ (Grant’s pro tip: AI never uses @ naturally, so it’s the perfect delimiter for Zapier loops).

2. Topic filtering — ChatGPT takes that raw list and narrows it down to the three ideas with the most viral potential for LinkedIn, again separated by @.

The outcome: a clean, filtered set of three post-worthy ideas that would have otherwise died as “we should post that” moments. Grant 1: blank page 0.

👉🏼 Grab the Topic Identification and Topic Filtering prompts to drop straight into ChatGPT

Step 2. “Make it spicy”🌶

Once the three filtered topics are ready, Zapier ships each into Claude with the original transcript. This is where raw ideas get teeth.

🎯 The “Make it spicy” prompt tells Claude to rewrite the idea as a short, assertive take and fold in any lived experience or data mentioned in the transcript. The result? A 1–2 sentence “spicy take” that actually sounds like something the founder would say. Even with AI in the loop, a distinctive point of view is what makes or breaks a post. Without it, you’re just adding to the LinkedIn noise.

From there, Zapier pipes each spicy take into a Slack channel. Every item arrives with context: the meeting source, the topic, and the draft itself.

The founder (Grant, in this case) can then:

• Approve or edit the spicy take directly in Slack

• Select the publishing platforms (LinkedIn in this case)

• Push it forward into ClickUp (or any project management tool) for scheduling

Grant added: “It was totally worth building an intermediate step to sample the idea before committing to a full post.”

🌶 Grab the “Make it spicy” prompt and paste it straight into ChatGPT or Claude to replicate this step.

3. Content generation

Once the founder approves the spicy take in Slack, it’s time to go from snippet to full post.

• AI’s job: Zapier moves the spicy take into Claude, which runs it through two layers. First, it applies tested hook and CTA frameworks to expand the snippet into a full draft. Then the Humanizer prompt reshapes it in a natural, conversational voice; contractions, varied sentence length, casual asides, even small “messy” touches. This keeps the post sounding like the founder, not a robot.

• Founder’s job: The draft pops up in Slack. The founder can do a quick edit, approve it, or reroute it to a different platform (LinkedIn, Twitter/X, blog).

• Social media team’s job: Once approved, the post lands in ClickUp.

The social media team steps in to:

• Adjust formatting

• Categorize posts (e.g., TOFU/MOFU/BOFU)

• Add visuals, carousels, or design elements

• Eventually repurpose to more platforms

This split keeps the founder out of the weeds while making sure nothing reads like it came straight from a bot. We liked Grant even more when he said, “Anything that leaves your company should always have a human in the loop.”

👉🏼 Steal the Humanizer prompt to make sure your AI drafts read like you, not like ChatGPT on autopilot.

Step 4. Schedule and measure

Now that the content is pitch-perfect, the social team sets a posting date in ClickUp. On that day, Slack pings a reminder, the post goes live on the chosen platforms, and the URL is logged back into ClickUp. Exactly 48 hours later, Slack resurfaces the link so metrics can be added and saved for analytics. The loop is closed: content isn’t just shipped, it’s also tracked.

And while you can automate publishing, pretty much any platform still favors human posting. If you have the bandwidth, hit ‘publish’ yourself and drop a comment or two to keep the algorithm gods happy.

🤖 Tools powering the Workflow

Grant runs Call-to-Content on a lean stack: Fathom for transcripts, Zapier for routing, Slack for review, and ClickUp for scheduling. For the LLMs, he splits duties. ChatGPT handles extraction and filtering, while Claude takes over the writing. If you only want to keep things simple, just stick with your LLM of choice. Or, if you’re curious, test the difference by running the other on a free plan. What makes this setup powerful isn’t the tool count but the way it’s wired together: smart prompts, clear use case, and a structure that turns a straightforward stack into a maximum-impact workflow.

🎢 Highs, lows and Workflow warnings

Like every workflow we feature, Call-to-Content has its sweet spots and its breaking points.

✅ What shines

• Consistency unlocked: reducing the time spent on each post from 30 minutes to 3–5 minutes with automated spicy takes as the draft.

• Distinctive POV baked in: posts sound like the founder because they’re drawn from transcripts, not generic AI fluff.

• Human-in-the-loop Slack review ensures content quality.

• Scale: one single call produces multiple content angles, creating a backlog of posts to batch and schedule.

❌ What doesn’t shine

• Overwhelm: early versions of the workflow generated 5+ posts per call. Grant ended up with 750 drafts waiting in ClickUp.

• AI still can’t nail images or carousels. Visuals need a human designer.

• Without clear filters, confidential or irrelevant points can sneak in. Approval rounds are crucial.

⚠️ Workflow warning

Hard pill to swallow: this workflow won’t magically create your (or your exec’s) point of view if there isn’t one. AI only amplifies what’s already there: deep knowledge, strong opinions, lived experience. The prompts help sharpen and scale, but they can’t invent substance. And that’s why the Slack review step matters. Skip it, and you’ll be shipping raw AI drafts that read like generic AI slop.

✨ The Goldflow

It’s telling when someone who automates for a living is the first to admit where it falls short. We’d tattoo Grant’s line on our foreheads: “Anything that leaves your company should always have a human in the loop.”

That’s the paradox of today’s workflow: it saves hours by killing the blank page, but its real strength is in the boundaries. No auto-posting, no AI-generated images, final judgment always human.

The other big lesson: founder time is leverage. One hour of conversation can fuel weeks of content. The goal isn’t to go viral, it’s to show up consistently with useful, opinionated takes. In a world of n8n bros flexing 100-step dream automations that will never run, the real power play is simple: keep it human, keep it consistent.

See you at the next Workflow,
Sara Stella & Diandra ✌🏼

🌶 If this workflow saved you from the blank page, pass it on.

Share

🌯 It’s a major wrap

This was the fourth edition of The Workflow, a bi-weekly newsletter by Sara Stella Lattanzio & Diandra Escobar, featuring one real workflow that moved the needle for someone even smarter than the system they built.

Subscribe now

Today’s guest is Lashay Lewis, aka the queen of bottom-of-funnel (BOFU) content. Two and a half years ago, she built the BOFU framework she wished she’d had back when she was grinding at a content agency.

That Notion-based dashboard morphed into Authority Plug, her specialized bottom-of-funnel content advisory for high-growth SaaS teams. As the AI boom hit, she saw an opportunity to scale client operations by building her own AI-powered platform to run her consultancy end-to-end.

Unfortunately, we can’t hand over her platform (off the market, sorry). But we can hand you this: a dead-simple starter kit for BOFU content that performs—even if you’re stuck with messy docs and one faithful LLM tab.

Subscribe now

✅ What you’ll learn

• How to build a BOFU content strategy with AI

• How to gather information to create high-quality product-led content

• How to run BOFU keyword research for Google + LLMs

• How to create product-led content briefs that convert

🧨 What triggered The Workflow

Creating bottom-of-funnel content that actually converts is harder than most marketers want to admit. You’re writing for buyers who are already weighing options, not scrolling for generic advice. They want proof, facts, specifics, and reasons to choose you over the competition.

“This is how most SaaS companies approach BOFU content: skim a few product facts, research some keywords, write something ‘optimized’, hit publish… then wonder why it doesn’t convert.” Lashay told us.

She has a different approach: her frameworks combine product messaging, sales data, and customer intelligence. When AI entered the picture, she layered automation on top to minimize the busywork.

“At some point I realized I wasn’t losing deals to other consultants. I was losing them to content tools”, she admitted. The systemization that made her work so effective also blurred the line between consultancy and SaaS.

That tension pushed her to rethink the balance between bot and machine. The result is her new consultancy BOFU.ai, which combines the expertise of a senior content specialist and the scalability of AI.

👇🏼 At the core is the proprietary workflow we’re about to break down step by step.

The impact: Interview prep to first draft took half the time compared to our usual pace. Admin work dropped by 70%. And we finally stopped chasing transcripts and started shaping insights.

Want to replicate this workflow or tweak it for your own use case? Keep reading.

🛠️ How to build The Workflow

Lashay runs her consultancy on a custom platform. You don’t need that. What you do need are the four core steps behind her BOFU strategy.

🎁 Each step includes a signature prompt you can drop into your favorite LLM to guide research and content. In Step 4, you’ll also get a ready-to-use BOFU brief template, the same structure Lashay now runs inside her platform.

Step 1. Collect product and customer documentation

Gone are the days when adding ‘x’ amount of keywords in the article will get the job done. No SEO tool can replace the research needed to ensure your content resonates with your buyers.

BOFU content is only as strong as the inputs behind it. That means pulling business-relevant insights that actually show how customers think, talk, and buy.

Gather these key docs to build a strong product-led content strategy:

• Sales decks, pitch materials, and product docs

• Sales call recordings or CRM notes (Gong, Chorus, etc.)

• FAQs, help center docs, and support logs

• Any doc where differentiators and objections are hiding

AWESOME. But without a system to distill and organize this knowledge, it stays buried. Create a list of pains, benefits, language, and buying context that drive conversions.

Fundamental questions you should answer:

• Who buys the product and why now?

• What pain triggered their search for a new solution?

• How do they describe your product’s value in their own words?

• What competitors do they mention on calls?

• What do they say your product helps them do better?

• What tools were they using before, and why did they switch?

This is where great product-led content starts: not with keywords, but with context.

🤖 How to speed up things with AI

Skip the manual slog: let your favorite LLM surface the insights that matter. AI can help you:

• Summarize large chunks of internal notes

• Extract patterns across different customer interviews or pain points

• Identify recurring objections or key differentiators

• Cluster quotes around goals, pain points, or feature mentions

• Highlight product benefits in the customer’s own language

💬 Grab the prompt to help you extract useful BOFU insights from your docs

Step 2. Map the competitive landscape

Once you’ve mapped the pains and the terminology your buyers use, it’s time to check what they are hearing from everyone else. Competitive research is the lifeblood of BOFU content.

“Every SaaS company says ‘Our product is the easiest to use.’ Cool. But if I read five competitors saying the exact same thing, it doesn’t help me make a decision. We need to give customers actual contrast.”

To do that, you need to articulate value propositions, key features, and capabilities clearly.

Here are the sources to analyze to run SaaS competitive research:

• Competitor websites, especially product pages, case studies, and comparison pages

• Review sites (G2, Capterra, TrustRadius)

• Customer tickets, reviews, and comments about competitors on sales calls

🤖 How to speed up things with AI

AI can help you research competitors, compare your product to theirs, and identify competitive whitespace for your product and brand strategy. Lashay recommends using Perplexity for discovery and ChatGPT/Claude for analysis.

💬 Grab the prompts to profile your competitive landscape and to build a comparative matrix

3. Run BOFU keyword research that goes deep

This isn’t your average SEO agency’s “sprinkle in some modifiers and pray for traffic” keyword strategy. “Customer research drives keyword research.” Period.

Lashay uses the BOFU Radar. This quadrant-shaped framework will help you identify high-intent terms across four angles:

BOFU keyword research isn’t just blogs and Google anymore. This exercise uncovers opportunities both in search results and LLM responses for the queries that matter:

• “Best X for Y” prompts (classic)

• Pain-driven queries (”I’m a compliance manager looking for an onboarding tool…”)

• Feature-level tool comparisons (”tools with built-in tax reporting”)

🤖 How to speed up things with AI

Once you’ve filled up the BOFU radar, let AI turn the insights you’ve collected into search terms and prompt-ready phrasing that actually convert.

💬 Grab the prompt to generate BOFU keywords and LLM-friendly queries

Step 4. Create conversion-worthy content briefs

Buyers can spot the difference between a “ChatGPT whipped this up in two minutes” article and a legit BOFU piece.

The first? A lifeless listicle.
The second? Structured, specific, and packed with facts, comparisons, and answers to real buyer objections.

This step is about turning your research into briefs that help writers transform raw intel into content that informs, differentiates, and actually converts.

What should a strong BOFU content brief include?

• Clear answers to real buyer objections

• Direct product comparisons backed by evidence

• Stats, quotes, and CTAs that guide readers toward conversion

• Content that matches high-intent, commercial queries
  

Use this BOFU content brief template to plug in your research and structure briefs that eliminate blank-page panic.

And yes, AI can help you jump from brief to draft fast, but we’ll save that for another issue. For now, nail the brief, amigos.

🤖 Tools powering the Workflow

Lashay runs BOFU.ai on a stitched-together stack hosted on Moonlit, combining research tools, automation, and documentation to keep her process running at scale. If this setup feels intimidating, that’s because it is. The good news is you don’t need a tech rocketship to launch your BOFU content strategy. Start by organizing your product, customer, and competitor insights in Notion. Then connect your favorite LLM directly through Notion MCP to analyze, extract, and cluster what matters.

🎢 Highs, lows and Workflow warnings

Lashay’s workflow delivers where off-the-shelf tools collapse. “There was no tool that could do what I needed. Asana couldn’t do it. Notion was the closest, but my clients didn’t even like editing in Notion. I literally had to build this.”

✅️What worked: her system saves both time and headcount. BOFU.ai consolidates the work of three specialized content and product roles into one repeatable engine. Saas companies plug directly into a process that ramps faster, aligns sales, product, marketing, and delivers pipeline.

❌What didn’t work: getting AI to nail nuance. “Finding that sweet spot between enough context and too much, where it starts to hallucinate… I haven’t nailed that for two years.” Automation powers speed, but clarity, structure, and style still demand a human brain. Or as Lashay puts it: “AI is the modifier, not the use case.”

✨ The Goldflow

Lashay’s biggest lesson from building her platform isn’t about prompts or integrating different tools. It’s about realizing she’s capable of way more than she gives herself credit for.

AI gave her a faster way to get thoughts out, systemize nuance, and operationalize what others would call “too messy to scale.” The brilliance is still hers. AI just helped her prove it.

Under the hood it took two years of trial, error, and plenty of frustration to get here. And it’s still not perfect; context, structure, and depth remain her biggest fights with AI. But the payoff is massive: a workflow that saves clients serious money, tears down team silos, and consistently produces BOFU content that converts.

Run it right and you’ll bottle nuance into a repeatable system. Run it sloppily and you’ll end up with yet another AI-written blog that reads like your FAQ page threw up.

See you at the next Workflow,
Sara Stella & Diandra ✌🏼

🔨 Know someone trying to nail their product-led content?

Share

🌯 It’s a major wrap

This was the fourth edition of The Workflow, a bi-weekly newsletter by Sara Stella Lattanzio & Diandra Escobar, featuring one real workflow that moved the needle for someone even smarter than the system they built.

Can’t survive two weeks without us? Catch up on past Workflows and steal the playbooks you missed.

✍️ The Research Report Workflow to slash production time by 50%

🎯 The Lead Scoring Workflow that doubles free-to-paid conversions

🔍 The Brand Positioning Workflow for the AI era

Subscribe now

Whether you work in marketing, sales, or run your own thing, each issue delivers a real-world workflow from a business pro that you can replicate as you read.

Last time, HeyReach’s CMO showed us the behind-the-scenes of an AI lead scoring Workflow that doubled their free-to-paid conversions. This week, we’re jumping from something extremely measurable like lead scoring to something typically impossible to define: brand positioning. Spoiler alert: this week's guest built a workflow to make it more tangible (without breaking the magic).

Subscribe now

✅ What you’ll learn

• How to build a complete brand positioning workflow with AI

• How to measure and quantify “brand” as much as possible

• How to run competitive brand analysis, fast

• How to create a reliable scoring system to analyze your brand

• How to craft resonant messaging around your key differentiators

🧨 What triggered The Workflow

Our guest, Clay Ostrom, founder of brand strategy agency Map & Fire, worked in the past with giants like Google, Sony, and Mattel and has spent the last 10 years helping everyone from scrappy startups to SMBs stand out in crowded markets.

Thanks to the AI boom, that job just got a whole lot harder.

“I started noticing how businesses, especially in SaaS, aren’t just up against two or three rivals anymore. They’re up against 10–15, each chipping away at market share.”, Clay told us.

The good old “gut feel” approach to positioning can’t keep up. Agencies like his feel it even more: every new client requires starting from zero on competitive analysis and messaging, often at high volume and under brutal timelines.

“It’s the classic ‘Here are 20 competitors and 50 internal docs to review for our kickoff tomorrow’ kind of thing. We need better tools. As competition levels go up, we have to fight fire with fire.”

His answer? A workflow that turns fuzzy positioning into something measurable and actionable, helping you quickly analyze dozens of brands, understand their superpowers, and position against them effectively.

Clay even built his own tool, SmokeLadder, to automate and scale the process. But here’s the best part: he’s sharing a slimmed down version of that workflow that you can run with any of your favorite LLMs (probably ChatGPT, Claude, or Perplexity, we’re guessing).

Want to replicate this workflow or tweak it for your own use case? Keep reading.

🛠️  How to build The Workflow

This Workflow is your playbook for defending (and winning) your brand budget.

It takes one of the most abstract, “fluffy” aspects of marketing — positioning — and makes it concrete, structured, and even measurable. Follow these steps and you’ll have hard evidence to back up your brand decisions, instead of hand-waving through strategy decks. If more SaaS companies did this, we’d all spend less time squinting at homepages wondering, “But what does this tool actually do?”

Clay has gifted us some simplified prompts for each step listed below (thank you, Sir!), so you can grab a coffee, fire up your favorite LLM, and start running your first AI-powered positioning sprint today. If you want to go further, SmokeLadder (his platform) pulls these steps into one streamlined workspace, provides deeper consistent scoring, and layers in benchmark data to help you see exactly where your brand stands.

💘Get all the prompts for steps 1–5 of the brand positioning Workflow and run them with whatever LLM you love 💘

1. Nail your brand fundamentals

Before you can position against competitors, you need a crystal-clear definition of your own brand’s foundation. This step is all about answering: Who do we serve? Where do we play? What needs are we solving?

To define your brand, you need absolute clarity on these points: 

• Target customer – the exact audience you’re trying to reach

• Target category – the market space you want to own

• Offering definition – what you actually provide

• Core benefits – the value your product or service delivers

💬 Clay doesn't beat around the bush: “If you can’t articulate this internally in plain language, it’s impossible to differentiate externally.”

For demonstration purposes, we'll use ChatGPT to analyze and build Tella's positioning, a screen recording and editing tool:

2. Analyze the brand category

Now it's time to understand the ecosystem you’re stepping into. This step is all about answering: “Who else plays in our space? How are they framing themselves? Where do we have room to stand out?”

You'll need to map out:

• Category alignment – the primary category your audience would place you in

• Alternative categories – other spaces you could credibly compete in

• Key competitors – the main players to benchmark against

• Challenger brands – emerging or niche players making interesting moves

🗺 By now, you should have a clear map of the players that matter most; the ones you need to contrast with if you want to stand out.

And that’s just the enrichment layer. With a detailed profile built out (including ICP match, firmographics, behavior, and job titles), each lead is now ready for scoring, assignment to a responsible team member, and for a suitable next step.

3. Score your brand (and your competitors)

This is where positioning gets tangible. Instead of playing the “We’re the best!” card, you will turn differentiation into numbers.

Clay’s philosophy: “If you can quantify it, even if it’s subjective, it’s a lot easier to find points of improvement.”

Score your brand and your competitors across these dimensions:

• Value delivered – how much your offer satisfies customer needs

• Differentiation – how clearly you stand apart in the market

• Category fit – how well you align with the space you’re competing in

🧮 The scores aren’t set in stone; they’re a reliable compass for spotting strengths, weaknesses, and areas to refine.

4. Lean into your brand differentiators

With scores and comparisons in hand, it’s time to figure out where you can stand out the most. This step breaks your brand into:

• Advantage points – the unique value you deliver that others can’t match

• Table stakes points – things you must have just to stay in the game

• Weakness points – areas where competitors currently outshine you

• Where to double down – the strengths that can set you apart even more

🤑 The goal here is simple: maximize separation from the competition by leaning into your most unique points of value. As Clay puts it, “The magic isn’t just spotting where you’re strong; it’s deciding where that strength will actually win you customers.”

5. Craft resonating brand messaging

Wild how many people think this is the first (and often the only) step of brand positioning. In reality, it’s the last mile. By the time you get here, you’ve already mapped your category, identified competitors, scored your positioning, and pinpointed your unique advantages. Now, you’re turning all that evidence into words.

In this stage, you’ll craft:

• Headline concepts – The core statement that captures your brand’s promise.

• Subheadline concepts – A concise expansion that explains how you deliver that promise.

• Customer pain points – The problems your audience feels most acutely.

• Brand benefits – The tangible value you deliver to solve those pains.

Clay notes, “You’re no longer starting from a blank page; you’re now starting from hard evidence.” That’s why your messaging will land stronger, read sharper, and connect faster.

⚠️ Note: While AI tools can help you brainstorm messaging ideas and angles, we still recommend having a human copywriter shape the final product because nuance, tone, and emotional resonance are what make the words stick.

🤖 Tools powering the Workflow

Sara made fun of her pretty basic tech stack for executing the Marketing Report Workflow, but this tech stack is next-level minimalist. All you really need is your LLM of choice (ChatGPT, Claude, Perplexity… pick your poison).

Alternatively, if you want an all-in-one option that lets you plug in URLs and get structured scores, visualized outputs, and strategic insights without juggling 15 spreadsheet tabs, that’s where SmokeLadder comes in. That's it.

We love workflows like this because they show how tech can accelerate an evergreen strategic project, making it faster, scalable, and easier, without pretending the tech is the strategy. Positioning still depends on your thinking and product, not on how many tools you stack.

⚠️The size of your tool list says nothing about how good of a marketer you are (although many people on LinkedIn clearly disagree).

🎢 Highs, lows and Workflow warnings

The magic of this workflow is how straightforward and repeatable it is. Once you’ve run it a few times, it becomes second nature. The best part is how it turns a notoriously fuzzy process into a clear, objective sequence you can run for any brand (or multiple brands at once, if you're an agency) without drowning in open tabs or spreadsheet hell.

Clay points out that having structured scores for positioning “pulls the soul of your brand out of your head and gets it on paper” — a huge advantage when conversations drift into subjective debates. Instead of arguing over gut feel, you’ve got data.

That said, here are the most common brand positioning mistakes to avoid:

• Garbage in, garbage out. If your inputs (competitor sites, messaging, product info) aren’t accurate or up to date, your scores will reflect that. Clay stresses the importance of sanity-checking sources before running the analysis.

• Don’t skip the foundation work. Many people want to jump straight to crafting messaging, but without clear answers to “Who do we serve?” and “What category do we play in?”, you’re polishing fluff.

• AI ≠ copywriter. Even with sharp insights and competitor breakdowns, the final messaging still needs a human brain (and tone). Think of this workflow as your research muscle, not your brand voice.

✨ The Goldflow 

AI is fast. Too fast. But the beauty of this workflow is that it forces you to slow down where it matters. Positioning isn’t a “set it and forget it” document you dig up once a year; it’s a living strategy that needs regular stress-testing. AI just makes it easier to run that test so it doesn’t feel like a dreaded maintenance chore.

Clay’s biggest takeaway is to stack prompts and break the process into pieces. Don’t feed an LLM the whole brand universe and expect gold to spill out. Run the competitive scan. Validate your sources. Score your position. Then step back and think. That’s the part AI can’t do for you (and the part that makes all the difference).

It’s tempting to chase the perfect headline straight out of the gate, but the real win here is having a repeatable, data-backed lens on your market. When you know exactly where you stand, every creative choice stops being a gamble and starts being a strategy.

Maybe you’ll even join the top 3% of B2B companies where visitors instantly “get” what you do.

See you at the next Workflow,
Sara Stella & Diandra

💡Know someone working on their positioning?

Share

🌯 It’s a major wrap

This was the third edition of The Workflow, a bi-weekly newsletter by Sara Lattanzio & Diandra Escobar, featuring one real workflow that moved the needle for someone even smarter than the system they built.

We could get used to this: collecting battle-tested workflows from ridiculously sharp people and passing them on like contraband. Next issue’s already cooking.

💌 Loved this issue and want to show us a little kindness?

Subscribe now

This week’s issue is special: after our launch edition, it’s our first-ever guest feature, and we'd never start with just anyone. We sat down with Vukasin Vukosavljevic, CMO at HeyReach, a Balkan SaaS that hit $8M ARR in just 26 months (and a good friend of ours), to break down how to build an AI-powered lead scoring system that can run your entire go-to-market (GTM) engine.

Subscribe now

✅ What you’ll learn

• How to score leads automatically using data enrichment and Clay tables

• How to route leads to CS, Sales, or Marketing instantly

• The stack powering The Workflow that doubled free-to-paid conversions

• How to build a system that evolves with your business

🧨 What triggered the Workflow

Lead scoring is one of those things every GTM team knows they should systematize. But most don’t. Why? Because building an automated, scalable lead scoring engine that’s accurate, flexible, and doesn’t collapse under the weight of conflicting opinions on what makes a qualified lead has always felt like a pipe dream.

Until now. At HeyReach, a LinkedIn outreach platform pulling in 2K+ free trial signups per month, Vuk and his team couldn’t afford to waste time chasing the wrong leads, or even worse, missing out on the right ones.

So they created a custom-built, automated lead scoring system that qualifies inbound leads using AI and routes them across sales, growth, marketing, and CS teams. It blends human logic, agentic AI, outbound campaigns, and a good ol’ human touch, exactly where it makes a difference.

Want to replicate this workflow or tweak it for your own use case? Keep reading.

🛠️  How to build the Workflow step-by-step

This lead scoring system has become the operational brain of HeyReach’s GTM machine. It’s smart, scalable, and something you should seriously consider, especially if you’re juggling high inbound volume from multiple sources like LinkedIn, YouTube, email, organic traffic, and referrals (key acquisition channels for HeyReach).

The impact: Conversion from free trial to paid jumped from 9% to 17%+. Salespeople now spend their time following up only with highly qualified leads. Marketing is crushing acquisition targets…and finally getting those high-fives from Sales. Lovely.

This exact workflow isn’t plug-and-play, but the logic is. Here's how it works.

1. Align on your ICP (and how to best serve them)

“We shouldn’t care about how many signups we get. We should care about how many qualified signups we get.” That insight was the real trigger for the workflow. Vuk and HeyReach’s CRO, Ilija Stojkovski, were frustrated by two things: a lack of clarity on which free trial sign-ups were actually ideal customers, and a relatively low conversion rate at the time (~9%) from trial to paid. 

“I wanted to know: Are our acquisition channels bringing in the right people? Are we spending time where it counts?” Vuk asked himself.

The first step is to align internally on customer tiers and their importance to the business. Without that, any system or tool will only cause even more problems.

“You can’t score leads if you don’t know what a good one looks like.” Vuk and the team mapped their ideal customer profile into tiers:

🤝🏻 Tier 1–2: High-fit, routed to CS or Sales for human onboarding

📩 Tier 3: Medium-fit, pushed to Marketing and self-serve flows

🚫 Disqualified: Not the right audience (e.g., ecomm agencies using non-LinkedIn channels)

Only now it’s the time to get nerdy with tools.

2. Enrich inbound leads using AI agents

At HeyReach, enrichment begins the second someone starts a free trial. The team collects only a business email at signup; no friction, no long forms. From there, automation takes over.

🧲 Step 1: Trigger

A new contact is added to HubSpot (their CRM). That’s the only input the system needs to get started.

🧠 Step 2: Enrichment

• Clay orchestrates a full enrichment sequence:

• Apollo fetches company and contact info 

• The system pulls website and LinkedIn data to understand the lead’s business model and ICP, assessing whether it would benefit from using HeyReach.

• If the trial signup is part of a team, only Admins and Power Users get enriched and contacted. All other team members are filtered out to avoid spamming the entire org (and to save enrichment credits).

• Custom prompts in Clay generate clean, properly formatted job titles to use for HeyReach’s outbound email campaigns (more on that later). 

All of this is handled in a single, monster Clay table (plus connected sub-tables), which acts as the central nervous system for scoring and routing leads.

And that’s just the enrichment layer. With a detailed profile built out (including ICP match, firmographics, behavior, and job titles), each lead is now ready for scoring, assignment to a responsible team member, and for a suitable next step.

3. Score and qualify leads (and act upon them!)

Up next: a maze of Clay columns (even Vuk admits some look like Chinese) handles scoring and lead routing automatically.

The system evaluates every free trial lead across multiple dimensions—company size, industry, ICP match, team composition, LinkedIn presence, and past tool usage. Each data point feeds into the logic that determines where the lead belongs and what action to take next, based on the four tiers defined in Step 1 (Tiers 1-3 and “Disqualified”).

This qualification is fully automated through several deeply connected Clay tables, each customized for Sales, CS, or Marketing. Teams focus only on the KPIs and signals that matter to them. All enriched and (dis)qualified contacts are synced back to HubSpot, which acts as a backend tracker while Clay remains the daily operational hub.

“This is the first company I’ve ever worked for that has their CRM under control. I don’t even need to log into HubSpot. The information lives in Clay anyway.”

Man, did he make us jealous with this statement.

4. Activate leads with targeted outreach

Once every lead is enriched, scored, and bucketed, it’s time to act. For Tier 1 and Tier 2 leads (the high-fit ones), the system triggers personalized follow-ups: assigning a CS owner in HubSpot and launching a multichannel outreach sequence using Smartlead for email and HeyReach for LinkedIn (so much for dogfooding!).

“It’s not just about qualification. It’s about knowing exactly what to do with each signup, without having to log into five tools to figure it out.”

Here’s a sample cold email template they use. Now you’ll see where all that ICP enrichment data they enrich inbound leads with really shines:

Hi [NAME],[CS NAME] here, your dedicated Account Manager. Think of me as Robin to your Batman. 💪

I’ve done some research and noticed your team is targeting [your lead’s ICP job titles].

If you’re up for it, I’d love to give you a personal walkthrough of [your tool] and help you build high-converting LinkedIn campaigns.

We could cover things like:

• Building GTM playbooks

• Sourcing leads from anywhere, capturing social signals, and enriching with any data points you need

• Connecting [your tool] with your GTM stack [specific tools the lead uses]

Want me to share some LinkedIn strategies that always work?

Happy hunting, [CS NAME]

On LinkedIn, each campaign branches based on whether a contact is already connected or not:

• 💬 If already connected → Send a message, engage with a recent post, and follow up if needed.

• 📬 If not connected → Send a connection request. If they accept, follow up with a message. If not, trigger an InMail.

The HeyReach API makes this seamless: campaign senders are dynamically assigned based on the lead’s HubSpot owner (if one exists), so they get messages from a familiar face. One campaign, multiple senders, all automated.

Vuk shared some results from a recent outbound campaign following this strategy:

🧑🏻‍💻 LinkedIn: ~47% connection rate and ~20% reply rate

📧 Email: ~12.5% reply rate (with zero optimization for opens or clicks)

Where can we sign to get that??

5. Keep refining

This thing doesn’t just run. It evolves. What started as a single lead scoring system quickly turned into a full-blown GTM intelligence engine. The team keeps layering new workflows on top: acquisition reports, action-based ICP intelligence, campaign tailoring, and routing to human-led or self-serve onboarding. “Next up, my dream would be to have a fully automated system for personalized onboarding, tailor-made retention routing, and what I call my Mount Everest: fully automated churn prevention.”

🤖 Tools powering the Workflow

Not the leanest stack, admittedly, but the results more than make up for it.

Clay is the brain, running enrichment, scoring, and routing logic across Sales, CS, and Marketing. Apollo plugs in verified emails and company data. HubSpot acts as the CRM backbone where enriched records and ICP tags are stored. Smartlead takes over for automated cold email outreach and retargeting, while HeyReach powers LinkedIn campaigns synced directly with HubSpot owners through Clay’s tiering logic.

Together, they form a GTM engine that qualifies and activates leads without the team juggling five dashboards.

📌 Disclaimer: you’ll probably need a tech-savvy friend (Vuk works with several automation magicians) to replicate something this intricate.

🌟 The good news? You can swap in your preferred tools and still build a similar system. Vuk even rebuilt the whole thing in Make, because apparently managing all of this wasn’t enough. So, whether you’re using Clay or Make, Smartlead or Instantly, HubSpot or Pipedrive, this workflow adapts if you’re willing to go deep.

🎢 Highs, lows and Workflow warnings

Like every workflow we feature, this one has its sweet spots and its breaking points.

✅ What shines
Clarity unlocked: “You’re not blind anymore,” as Vuk puts it. The team stopped guessing and started knowing — what channels convert, which leads deserve attention, and why certain ICPs churn.

Conversion doubled: Within a quarter of launch, free-to-paid rates jumped from 9% to 17%.

One source of truth: Sales, CS, and Marketing now work from the same enriched data without tool-hopping, conflicting reports, or duplicate effort.

More bandwidth: With manual lead qualification reduced, the team spends time on GTM strategy instead of admin bottlenecks.

❌ What doesn’t
Maintenance load: Prompts, syncing across tools, and refining ICP logic demand constant iteration.

Complexity tax: Scaling the logic required collaboration across departments and plenty of “techie sparring partners.”

⚠️ Workflow warning
The system solves qualification and alignment, but it won’t erase every CMO headache. As Vuk joked, “Some days I still feel like I’m crying over product marketing.”

✨ The Goldflow 

We close every interview with the same question: "What did building this workflow teach you about AI and automation on a macro level?"

Vuk didn’t hesitate. "My delegation skills are becoming crazier. Now I can delegate stuff to AI agents. I don't need to do it anymore on my own."

He’s not just talking about task automation, but REAL delegation. To agents, logic flows, and systems that just do the thing without human intervention.

And because Vuk’s the kind of person who loves solving problems that used to keep him up at night, seeing them finally melt away feels like magic. "It sounds fluffy, but it gives me that feeling of: Okay, now we can actually do the things I used to see only in movies."

Because the entire point of building systems like this is one: remove the mental load of hamster-wheel admin work and chaos.

Because yeah, you might still cry over product marketing, but at least now your workflows cry with you. And they file the right lead under "Tier 1" while doing it.

See you in two weeks,

Sara Stella & Diandra

😱 Know someone struggling to set up a lead qualification system?

Share

🌯 It’s a major wrap

This was the first-ever guest edition of The Workflow, a bi-weekly newsletter by Sara Stella Lattanzio & Diandra Escobar, featuring one real workflow that moved the needle for someone even smarter than the system they built.

If you missed the first edition (where Sara Stella built a full AI-powered editorial engine while yelling “AI isn’t the strategy” from the rooftops), go catch up. It might include traces of desk research that doesn’t melt your brain, outreach that doesn’t suck, and transcripts that file themselves.

Subscribe now

Whether you work in marketing, sales, or run your own thing, each issue delivers a real-world workflow from a business pro that you can replicate as you read. 

This is our very first issue, and for the special occasion, we kept things in-house.
I let Diandra interview me. You’re getting a behind-the-scenes look at the exact workflow I used to create a qualitative marketing research study using AI tools like ChatGPT, Zapier, and Fathom.

Starting from the next issue, you’ll meet our real guests: operators, marketers, and builders we actually look up to — the ones creating workflows worth stealing.

Let’s kick this off, shall we?

Subscribe now

✅ What you’ll learn

• How to plan and create B2B research reports

• How to recruit and reach out to study participants

• How to automate a huge chunk of admin work

• How to use ChatGPT to cluster quotes and extract themes

• How to slash overall research report production time with AI

🧨 What triggered The Workflow

Research reports are one of the top B2B content formats, but an ABSOLUTE nightmare to produce. You typically face two bad options: outsource it to an agency charging five figures for a glorified PDF, or keep it in-house and watch it drain your team and block the content pipeline for months.

Over the years, across different roles, I’ve been refining a better way. And I finally put it to the test for a recent project.

My team and I set out to create a qualitative industry report to explore key industry trends as a core piece of our quarterly content strategy. The starting point: ~40 expert interviews to run, a tight timeline, and a lean crew (leadership, writer, designer, and me as project lead).

None of us had the time to waste on admin hell, so we built a workflow that turned every interview into structured, shareable data with minimal friction. No more chasing transcripts, no more chaos folders. One teammate put it best: “It feels like a CRM for thoughts.”

The impact: Interview prep to first draft took half the time compared to our usual pace. Admin work dropped by 70%. And we finally stopped chasing transcripts and started shaping insights.

Want to replicate this workflow or tweak it for your own use case? Keep reading.

🛠️ How to build the Workflow step-by-step

I shout it from the rooftops in every single LinkedIn post of mine: AI and automation are NOT the strategy. And they’re definitely not a shortcut to avoid thinking.

I used them to kill the admin work so my team could focus on insights, not logistics. A crucial move when you don’t have a 50-person marketing department.

Replicate my workflow right away using whatever tool alternatives you’ve got:

1. Set the strategy, the old-school way

Before touching any AI tools, you’ll want to start the old-school way with an interactive team workshop in Miro to align on your goals, define your audience, and shape the narrative your report will tell. For us, the outcomes were clear: understand key industry trends through expert interviews, build relationships in the space, and position the brand as a thought leader in a relatively new vertical.

Once your point of view is locked in, design the study, define a study participant ICP, and craft a questionnaire that reflects your hypotheses and that can be easily adjusted to different interview profiles according to their specific focus.

2. Collect insights: deep REsearch

Next, kick off your desk research to map the landscape. But skip the classic 40-tabs-open chaos. Instead, use ChatGPT to synthesize background reports, surface market trends, and scan existing thought leadership. You’ll get a clear view of what’s already out there and where your report can actually add something new, without wasting hours in browser hell.

3. Find experts to interview

To recruit your study participants, take a hybrid approach. Start with warm intros from your network. No surprise here: they convert like crazy (we hit a 70% yes rate). But since that doesn’t scale, layer in cold outreach: build a Sales Navigator list, enrich it with Clay, and automate outreach via a tool like Dripify. We got a 40% “yes” on that. Not bad.

Send every expert a skimmable, well-designed one-pager that explains the study and include a Calendly link to book their interview slot. No endless email ping-pong.

4. Prep and personalize interviews

To avoid prep stress, send guests a personalized interview questionnaire five days before the call, tailored to their background and expertise. Ask unique questions if you want unique answers. Most interviewees will fall into 2–4 categories, so building a few tailored question sets already goes a long way. Want to take it up a notch? Build an automation that scrapes LinkedIn data and customizes each questionnaire accordingly (story for another newsletter issue).

And don’t forget to activate reminders in your scheduling tool because chasing confirmations or dealing with reschedulings manually should be a hard pass.

5. Record and transcribe the call

We ran all interviews on Google Meet, with Fathom handling real-time recording and transcription. Choose a tool that doesn’t just capture words, but actually highlights key takeaways so even the longest interviews stay scannable, killing the need to rewatch or hunt for quotes later.

6. Automation magic: post-call distribution

Set up a Zapier automation to save an insane amount of time. Use finalized interview transcripts as a trigger to automatically:

→ Update a Google Sheet with the interviewer’s name, summary, and transcript link.

→ Send a Slack notification so your writer can jump in right away.

→ Add the full transcript to a rolling Google Doc, where every interview lives in one place.

No clicking around. No chasing links. Just clean, centralized data—a marketer’s and writer’s dream!

7. Cluster insights with AI

Once you’ve gathered your interview transcripts, use ChatGPT to analyze qualitative data: identify themes, group expert quotes by topic, and summarize takeaways. This approach transforms hours of expert interviews into actionable research insights.

🧠 Pro tip: Create a dedicated project in ChatGPT where you can upload all your project docs, briefings, tone of voice, and branding guidelines in one place. It’ll save you hours in back-and-forth.

Here’s a prompt you can use (and I REALLY recommend stacking one request after another for best results):

I’ve uploaded multiple documents containing full-length interviews and project notes for a research report. Please use all of them to:

1. Identify up to 7 recurring themes and insights across the interviews.
   

2. Group key quotes under each theme (attribute by interviewee full name).
   

3. Summarize 3–5 key takeaways for each theme in clear, insightful bullets.
   

4. Write a concise executive summary that captures the big-picture insights from the research (max 6–8 sentences).

Please refer to the Brand guidelines document I’ve attached for writing style and TOV guidance. Let me know if you have any remarks, complaints or doubts before you get started.

8. Final steps: build the story and get feedback

With themes and clusters mapped, review the distilled document with everyone who needs a say. Refine the framing, narrative, and pick standout quotes. This tight feedback loop keeps the focus on the story, not the spreadsheet. From there, your writer can dive into the first draft and kick off the review cycle.

🤖 Tools powering the Workflow

My tool stack doesn’t win any awards for complexity or exclusivity (I’ll survive), but it dramatically cut the effort needed to ship this report. Here are a few more details about the tools I used, but feel free to swap in your preferred alternatives.

Fathom is the heart of the workflow: it transcribes interviews in real time, tags key insights, and surfaces actually usable quotes. But its value doesn’t stop at writing support. Each finalized transcript triggers a series of Zaps updating the interview tracker, pinging the project Slack channel, and dropping the transcript into a rolling Google Doc. “Where’s the transcript?” and “Did this interview already happen?” became questions of the past.

Clay and Dripify powered the cold outreach engine, helping us zero in on best-fit study participants with minimal manual effort. And once the transcripts started piling up, ChatGPT made sense of the chaos, clustering quotes, surfacing themes, and extracting takeaways fast.

And then there’s the “boring” but reliable foundation: Google Suite. Docs for transcripts, Sheets for lists and the interview tracker, and Meet for video calls.

🎢 Highs, lows and Workflow warnings

The biggest wins of implementing this workflow for me were saved time and the spotless, scalable organization of files, usually a magnet for human error. Down the line, our writer could start from curated, structured material instead of raw chaos. That alone saved at least six hours per week in admin work and cut our concept-to-draft cycle time in half.

Everyone always knew where the project stood, in real time. That made it easy to spot and fix bottlenecks (like a slow interview booking rate) before they snowballed. Because yes, it’s critical to save time on tasks that destroy value but it’s just as important to protect time for the ones that create it. I strongly recommend a weekly standup to align, unblock, and gather input from the team. It catches what async updates miss and helps keep motivation up when the project starts dragging.

Of course, tech isn’t a silver bullet. Fathom missed a few recordings entirely. It was active during the call, but the transcript never appeared and I only realized it after the guest had already signed off. If you want to avoid awkward follow-ups asking for notes, turn on Google’s built-in transcription feature as a backup. Just in case.

Another hiccup? Zapier’s task cap. We hit it midway through the project while I was on vacation (classic!), and the entire workflow just froze. Lesson learned: check your task meter early and understand what your setup consumes. The easiest solution? Turn on pay-per-task billing. Zapier will charge you per extra task once you max out your plan, so you never have to worry about it again.

✨ The Goldflow

Sue me: I haven’t tried all the other LLMs before deciding to use ChatGPT for this one. Not Claude, not Perplexity. Same goes for most of the tools in this stack. I just worked with what I had and what I knew best. I only start hunting for new tools when I hit a real cap… not to optimize something from 97% to 99%.

That’s the real tea. ☕️

It might go against the grain of how most people approach tooling today. There’s almost a stigma if you haven’t tested every AI tool under the sun, compared every feature, and stacked up 50 subscriptions like digital Pokémon. But here’s the truth: most companies already have enterprise tools in place, so you’re not starting from zero. And unless you’re solving the same complex, niche use case on repeat, you probably don’t need to build a rocket-ship tech stack.

See you at the next Workflow,

Sara Stella

❤️ How did you like our first issue? Let us know!

Leave a comment

🌯 It’s a major wrap

This was the first edition of The Workflow, a bi-weekly newsletter by Sara Stella Lattanzio & Diandra Escobar, featuring one real workflow that moved the needle for someone even smarter than the system they built.

See you in two weeks with another playbook worth stealing.

Subscribe now

If you’re here, it means you probably feel the same way we do about the current state of AI:

The internet is drowning in hot takes from people who became “AI experts” overnight, plus endless “revolutionary” automation hacks that collapse the moment you try to build them. A thousand tools, a thousand features, hype levels through the roof, and still no trusted place to actually learn how to use AI in a way that improves your skills and your work.

🛠️ So we decided to build that trusted source.

The Workflow is a Substack newsletter by Sara Lattanzio & Diandra Escobar that breaks down one proven AI-powered marketing and GTM workflow in each issue.

We show exactly how it’s built, what works, and what doesn’t, so readers walk away with prompts, tools, and steps they can steal right away.

We started The Workflow because the internet didn’t need another AI “10 tools that will change your life” hype thread. It needed real systems, battle-tested by people who’ve done the work, not just posted about it.

Subscribe now

🚀 What makes The Workflow worth your time

• Practitioners first – Not just big names. People who get their hands dirty.

• Honest AI commentary – What actually works. What doesn’t. What’s totally made up.

• One killer play per issue – No fluff. No 18-section brain fog.

• Tool discovery through context – We don’t list tools. We show how they work in practice.

🧠 Real practitioners. Not influencers.
We feature people who build things that actually work — not things that just perform well on LinkedIn.
No growth hype. No AI snake oil. No “I built this in one weekend and hit $12M ARR” nonsense. We balance big names with underrated operators from around the world — no echo chambers, no recycled guests, just people who actually do the work.

🔍 Battle-tested workflows only
Every play we feature has been used — in the wild, under pressure, with real outcomes.
We walk through it step by step, tool by tool. No vague “just set up a Zap” advice. No assumptions that you’re a prompt-engineering savant.

🤖 Automation with purpose
We’re not here to idolize AI. We’re here to show when it works, when it fails, and when doing things manually is still the smarter choice. We solve real problems. And we show how.

✂️ Short. Tactical. Immediately useful.
This isn’t a 12-scroll Notion board or a newsletter that takes longer to read than to implement.
You’ll walk away with one workflow you can actually use this week — no 50-tab research rabbit hole required.

🤝 Curious, not condescending
You shouldn’t need a PhD in prompt engineering or “AI” in your job title to understand these workflows.
We ask the questions you’re afraid to admit you have, and we answer them simply, clearly, and with zero shame.

🧭 Anti-buzzword. Pro-outcome.
We don’t care if it’s built with ChatGPT or a shared Google Sheet — if it drives results and saves your team’s sanity, it belongs here.

Subscribe now

💡 How you can help shape The Workflow

This isn't a one-way street. We're always on the lookout for practitioners who are quietly doing incredible work. The kind of work that actually moves needles, not just engagement metrics.

Comment on this post and tell us:

• What workflow challenges are keeping you up at night?

• Who's someone we should ABSOLUTELY interview? (Bonus points if they're not already on every other podcast)

• Got a workflow that's working magic for your team? We'd love to hear about it.

Your answers don't just help us plan content; they could easily become our next issue.

Leave a comment

Here's what happens next: Subscribe above to get the first issue straight to your inbox very, very soon. Each issue of The Workflow breaks down one battle-tested GTM workflow, soup to nuts, so you can steal it, adapt it, and make it yours.

Do you know someone who knows someone who would love this? Share it with them!

See you soon,

Sara Stella & Diandra

Share The Workflow