👋 Hey friends,

This week’s issue is a special one.

It started with a simple question I kept hearing from founders:
“How do you actually start an AI company that lasts?”

Not with another Copilot, not another idea generator — but with something that compounds into a category.

That question led me to the same insight shared by every breakout founder from Harvey to EvenUp:
You don’t start with a product.
You start with a workflow.

Because the best AI startups aren’t chasing innovation — they’re chasing friction.

So today’s edition breaks down how great companies build from one painful, repeatable workflow — and turn it into an empire.

Here’s what we’ll explore together;

The Wedge Playbook — Why every enduring AI company starts by solving one specific workflow that’s painful enough to scale.
The Founder Playbook — How to uncover and validate your wedge by observing real operators in action.
The Anatomy of a Perfect Workflow — The 5 traits that predict whether your workflow can turn into a product.
The Expansion Loop — How to grow from one workflow to an entire category, naturally.
Case Studies — Harvey, EvenUp, and Hippocratic — what they did differently, and what you can learn from them.

This one’s not about building faster — it’s about building truer.

Because when everyone else is chasing big ideas, the founders who win are the ones who find clarity in the small ones.

Let’s dive in.

— Naseema Perveen

Your product is only as good as the data it’s built on. Outdated, blocked, or missing web sources force your team to fix infrastructure instead of delivering new features.

Bright Data connects your AI agents to public web data in real time with reliable APIs. That means you spend less time on maintenance and more time building. No more chasing after unexpected failures or mismatches your agents get the data they need, when they need it.

Teams using Bright Data consistently deliver stable and predictable products, accelerate feature development, and unlock new opportunities with continuous, unblocked web access.

Power Your AI

 AI startups used to sound like this: 

“We’re building AI for finance.” 
“We’re automating healthcare.” 
“We’re rethinking education.” 

Today, those sentences raise red flags. 

The market doesn’t reward broad ambition — it rewards specific, repeated pain. 

Think of a “workflow” as a tiny factory of decisions that happens the same way hundreds of times a week. 

Each is a loop of micro-decisions that burns hours and bleeds money — and that’s where AI thrives. 

Workflows are predictable. 
They generate data. 
They repeat. 
And repetition is what AI eats for breakfast. 

Most AI founders begin by asking, “What can we automate?” 
The better question is: “What system can we rewire?” 

That’s the quiet shift happening in the most effective teams today. The early AI boom was all about task-level automation — writing emails, summarizing notes, generating content. But that was just the warm-up. The real transformation starts when AI moves from completing isolated tasks to running end-to-end workflows. 

Nick Scavone, CEO of Seam AI, explains it best: 

In other words, a small team with AI-driven workflows now operates with the leverage of an org ten times its size. 

Here’s the key difference: 

Rachel Woods, former Meta data scientist and founder of DiviUp, calls this “scaling what your team does best and unlocking infinite time for the work only humans can do.” 

The founders who win in this transition don’t just automate. 
They operationalize — turning AI from a helpful assistant into a core operating system for their business. 

Rachel Woods’ CRAFT Cycle framework shows what it means to operationalize AI the right way: 

1️⃣ Clear Picture — Define your process, who’s involved, and what success looks like. 
2️⃣ Realistic Design — Build the minimum viable automation that adds real value. 
3️⃣ AI-ify — Integrate AI into each step with context and data, not prompts. 
4️⃣ Feedback — Test, refine, and collect learnings continuously. 
5️⃣ Team Rollout — Train users, assign owners, and measure outcomes. 

This process turns messy, human workflows into stable, measurable systems — where AI executes, humans review, and both improve together. 

It’s not about coding models; it’s about teaching the organization to think in systems. 

Most AI startups fail because they chase outputs instead of operations. 
They automate one piece of a problem but never redesign how the pieces connect. 

That’s why “workflow-first” companies — like Harvey in legal or EvenUp in finance — scale faster. They don’t just replace effort; they replace inefficiency. 

When you automate systems, not tasks, three things happen: 

And that’s the new kind of leverage AI offers: not more output, but organizational intelligence that compounds. 

 Source : MicKinsey & Co.

When you zoom out across the most successful AI startups, a clear 3-step pattern emerges. But behind that simplicity lies a deeper operational truth: most companies don’t fail because they lack technology — they fail because they automate in the wrong order. 

78% of organizations now use generative AI. Yet 80% say it hasn’t improved earnings. 
That’s the AI paradox — high adoption, low impact. 

The difference between hype and results almost always comes down to this: 
The winners build around workflows. The rest build around ideas. 

Every enduring AI company begins with a narrow, high-friction workflow — something repetitive, measurable, and emotionally frustrating. 

Think Harvey starting with M&A contract review. Hippocratic tackling nurse triage. EvenUp focusing solely on claims calculations. 

These aren’t grand visions; they’re single pain loops. But by solving one deeply broken workflow, each unlocked compounding data, insight, and trust. 

This aligns with what operational experts call AI workflow selection — focusing on high-impact, technically feasible, and customer-facing processes. These are the loops where a 10% efficiency gain feels like a breakthrough. 

Examples include: 

Every workflow shares the same DNA: frequency, friction, structure, and stakes. 

Most companies use AI to automate tasks. The best use it to rebuild systems. 

That’s what Rachel Woods calls the shift from tasks to systems — turning isolated actions into self-improving workflows. 

Teams that operationalize AI this way use frameworks like the CRAFT Cycle: 

Clear Picture → Realistic Design → AI-ify → Feedback → Team Rollout 

It’s not about shipping models fast; it’s about creating tight feedback loops where humans refine and AI compounds. 

Successful founders treat early use cases like living organisms — observing, adjusting, retraining, and expanding only when the workflow runs cleanly from end to end. 

This structured obsession turns one process into a platform. 

Every great AI company grows through pull, not push. 

Once users trust your system to automate one workflow reliably, they’ll naturally ask for adjacent ones. 

EvenUp went from claims calculation → case building → full legal ops. 
Hippocratic went from triage → diagnostics → patient summaries. 

The pattern isn’t accidental — it’s the vertical wedge strategy in motion: 
Start narrow, scale wide — in the order reality allows. 

And the smartest founders prototype before scaling. They use no-code tools like Make, Power Automate, or Zapier to test if an automation actually works before committing to costly development. 

This lightweight experimentation validates assumptions early, exposes workflow edge cases, and builds confidence across teams. 

The Takeaway: 

AI’s future isn’t decided by who builds faster. 
It’s decided by who builds truer. 

The companies that win don’t automate everything — they automate one thing beautifully, then let that precision compound. 

That’s how a single workflow becomes a moat. 

If you’re building an AI company, your workflow is your market wedge. 
But not every process deserves automation — only the ones that create a flywheel when you fix them. 

Here’s what makes a workflow worth betting your company on: 

1️⃣ Frequency — The Repetition Advantage 

The best workflows are repetitive. 
They happen dozens, hundreds, or even thousands of times a month. 
That repetition gives AI what it craves most — data density. 

Each loop becomes a lesson, each mistake a micro-improvement. 
Over time, the system doesn’t just automate — it adapts. 
High-frequency loops build compound intelligence faster than any training dataset. 

Ask yourself: If this runs every week, how much smarter could it get in six months? 

2️⃣ Friction — The Emotional Signal 

Friction isn’t just inefficiency — it’s energy trapped inside a system. 
It’s the moment someone sighs, opens another spreadsheet, or manually copies text from one tool to another. 

That emotional fatigue signals unmet value. 
When a workflow feels mentally draining, not just time-consuming, users are ready to pay for relief. 

In other words: 
Pain is your product roadmap. 

3️⃣ Structure — The Learnability Layer 

AI needs structure like engines need fuel. 
Without clear inputs, outputs, and consistent logic, models can’t generalize. 

The perfect workflow has patterned data and predictable rules — even if they’re messy around the edges. 
It’s easier to layer intelligence on a structured system than to impose order on chaos. 

Start where there’s just enough consistency for AI to latch onto — logs, forms, checklists, or repeatable templates. 

4️⃣ Stake — The Impact Multiplier 

If it goes wrong, it hurts — financially or reputationally. 
That’s what makes it worth solving. 

AI thrives where the cost of error is visible. 
Whether it’s mispricing a deal, missing a compliance deadline, or delaying a claim approval, these are the loops where reliability equals revenue. 

Start where mistakes are expensive and time magnifies the risk. 

5️⃣ Expansion Potential — The Wedge Effect 

Every great workflow is a gateway. 
Fix one, and you expose the next five. 

Automating claims review naturally leads to case preparation. 
Streamlining triage unlocks diagnostics. 
Simplifying procurement opens analytics. 

Each solved workflow becomes a data and distribution asset — a foothold to enter adjacent markets with minimal resistance. 

The 4-out-of-5 Rule 

If your target workflow checks at least four of these boxes — frequency, friction, structure, stake, and expansion — you’re not just building a feature. 
You’re building a foundation. 

Because the perfect workflow doesn’t just scale your company. 
It compounds your intelligence. 

Let’s zoom out. 

Here are the industries where workflows are so broken, AI feels inevitable: 

🏥 Healthcare 

⚖️ Legal 

💸 Finance 

🏗 Manufacturing & Logistics 

These aren’t “tech” problems — they’re workflow bottlenecks that keep trillion-dollar industries frozen in 2003. 

AI is finally giving them a reason to move. 

Every founder dreams of building an empire. 
But the empires that last start with a single workflow — small enough to finish, painful enough to matter, and structured enough to scale. 

This is the Wedge Playbook — the framework behind how vertical AI startups like Harvey, EvenUp, and Hippocratic go from solving one workflow to dominating an entire category. 

Step 1 — Shadow the Work 

Before you build, observe. 

Spend a week watching someone do the task you want to automate — unfiltered, in their real environment. 
You’ll find the signal in their silence: 

These invisible moments are your product roadmap. 
They reveal friction that no survey or dataset can. 

When Harvey shadowed lawyers, they didn’t build around what lawyers said they needed — they built around what they actually did: scanning, comparing, highlighting, rechecking. 
That observation became their wedge. 

Step 2 — Build a Copilot, Not a Replacement 

The fastest way to build trust is to assist, not replace. 

Humans don’t want to be automated away — they want to be amplified. 
So design your product to make users faster, not vanish. 

Early copilots work best when they do the pre-work: suggesting, flagging, summarizing. 

They handle the 80% grind, leaving humans to handle the 20% that requires intuition, tone, or judgment. 

As Rachel Woods puts it: 

“AI doesn’t take over the work — it gives internal users technical superpowers.” 

When users feel supported, not sidelined, adoption compounds naturally. 

Step 3 — Mirror Their Mental Model 

Every workflow has a hidden logic — an unspoken decision tree inside the expert’s head. 
Your job is to map it. 

Ask users: 

This process builds what’s called synthetic intuition — where your model learns not just what to do, but how to think about doing it. 

If AI ignores human judgment, it breaks trust. 
If it mirrors it, it becomes indispensable. 

Step 4 — Expand via Adjacency 

Once you own one workflow, don’t pivot — expand horizontally from the inside. 

Follow your data trail. 
Users will start saying, “Can you do this part too?” 
That’s your signal for adjacency. 

It’s how EvenUp grew from processing claims → to building full case files → to managing entire legal ops. 
Each move wasn’t a new bet — it was a natural extension of trust. 

In the wedge model, the next product is always hidden inside the current one. 

Step 5 — Build a Moat of Feedback 

Your users are your best engineers. 
Every correction, rejection, or rewording they give you is free labeled data. 

Build loops that capture it automatically. 
Use those micro-corrections to fine-tune your logic, expand edge cases, and reinforce reliability. 

That’s how founders turn feedback into a data moat — one that compounds faster than any funding round. 

When your users train your product by simply using it, you’ve crossed the threshold from workflow → system. 

Start with one workflow that repeats, hurts, and matters. 
Design for empathy, not ego. 
Grow through adjacency, not ambition. 
And build a system that learns faster than your competitors can ship. 

That’s the wedge that becomes a category. 

Harvey (Legal) 

Started with one use case: contract review. 
Within a year, expanded into research and case prep. 
Their moat? Deep understanding of legal reasoning — not UI. 

Hippocratic (Healthcare) 

Focused on nurse triage and documentation. 
Scaled into diagnostic reasoning, powered by structured medical data. 
Their wedge was empathy — automating care, not just process. 

Each started with one boring workflow and turned it into a new category. 

How to Go From Curiosity → Clarity → Category Entry 

If you look at every breakout AI startup — from Harvey to EvenUp — they didn’t start with a “big idea.” 
They started with a workflow that broke humans a little bit every day. 

Here’s how to find yours: 

1️⃣ Pick an Industry, Not a Product Idea 

The best founders don’t chase ideas — they chase inefficiencies. 
Instead of starting with “I want to build something in AI,” start with: 

“Which industry runs on repetition, regulation, and frustration?” 
Healthcare, law, logistics, finance — these are markets built on structured chaos. 
The friction is baked in. And friction is where AI thrives. 

2️⃣ Interview Five Operators 

Forget surveys — go talk to people who live the pain. 
Ask one question: 

“What’s the task you dread doing twice a week?” 
Because frequency compounds frustration, and frustration predicts willingness to pay. 
The next billion-dollar product will come from a “dreaded Tuesday task,” not a sexy moonshot. 

3️⃣ Observe the Real Workflow 

People say what they think they do — but systems reveal what they actually do. 
Sit beside them (or on Zoom). Watch where they pause, hesitate, or cross-check. 
That’s where mental friction hides — and where automation can add real leverage. 
I learns best not from instructions, but from patterns of struggle. 

4️⃣ Recreate the Workflow in a Sandbox 

Before writing code, simulate the workflow in a spreadsheet, Notion, or ChatGPT. 
Use dummy data, replicate every step, and ask: 

“Where does this process break when humans step out?” 
This exercise makes invisible dependencies visible — and shows where humans still matter most. 
The best founders build tools that collaborate with humans, not compete with them. 

5️⃣ Launch Fast, Stay Narrow 

You don’t need 1,000 users. You need 10 who can’t live without it. 
Ship an MVP that solves one workflow end-to-end — then expand only when your users start saying, 

“Can you also do this part?” 
That’s pull, not push. 
And pull is how wedges become platforms. 

The Takeaway 

The next generation of founders won’t build “AI startups.” 
They’ll build workflow companies — invisible systems that quietly rewire how industries operate. 
If you can fix one workflow deeply enough, you won’t need to chase scale. 
Scale will chase you. 

Here’s what’s quietly happening across industries: 

In the 2010s, startups digitized workflows. 
In the 2020s, they automated workflows. 
Now, AI is interpreting workflows. 

The winners won’t be those who build faster, but those who build truer — AI that understands how real people think and work. 

Every workflow you fix becomes a data engine, every data engine becomes a product, and every product becomes a platform. 

That’s the new compounding loop of the AI economy. 

When you choose one workflow, it can feel limiting — almost small.
But what you’re really choosing is focus with leverage.

The paradox of modern AI startups is that the narrower your aperture, the deeper your moat becomes. Because scale doesn’t come from doing more — it comes from understanding one thing better than anyone else on earth.

The best founders don’t chase breadth; they chase depth.
They know every great system starts as a single solved workflow — a clean feedback loop that compounds clarity faster than competitors can copy it.

AI isn’t creating new industries.
It’s exposing the inefficiencies hiding in old ones.
Healthcare didn’t need new data — it needed new reasoning.
Law didn’t need new tools — it needed new trust.
Finance didn’t need more speed — it needed smarter sequence.

When you find that one broken loop — that everyday frustration buried inside bureaucracy — you’ve found your wedge.
Because fixing it doesn’t just save time.
It rewires how value moves through an entire system.

So while everyone else is chasing the next big thing, your job is simpler:
Find one small workflow that quietly touches everything.
Make it work so elegantly that the rest of the market reorganizes around you.

In an age of acceleration, clarity compounds faster than code.
And the founders who master that discipline — who build slower, truer, and deeper — will define what “scale” means in the AI era.

— Naseema

That’s all for now. And, thanks for staying with us. If you have specific feedback, please let us know by leaving a comment or emailing us. We are here to serve you!

Join 130k+ AI and Data enthusiasts by subscribing to our LinkedIn page.

Become a sponsor of our next newsletter and connect with industry leaders and innovators.