Vertical AI Startup: How Founders Find PMF

TL;DR: A vertical AI startup wins by solving one industry's painful, specific workflow so well that customers tolerate an unfinished product. Across 200+ PMF Show interviews, founders building vertical AI found product-market fit fast when the pain was acute — one hit $1M ARR in four months, another saw 40% cold-outreach-to-demo conversion. The trap is mistaking a viral horizontal launch for true vertical fit.

After interviewing 200+ founders on the PMF Show, the clearest emerging playbook is the vertical AI startup: applying AI to a single industry's hardest workflow rather than building a general-purpose tool. The founders who win do it by going deep on one painful problem, shipping something "barely there," and watching whether customers go out of their way to adopt it anyway. Here's how five founders approached vertical AI and what separated real product-market fit from false signals — with the numbers behind each.

What makes a vertical AI startup find PMF fast?

The strongest signal in vertical AI is customers adopting an incomplete product because the pain is so severe. Michel Tricot, co-founder of Airbyte, saw exactly this when his bare-bones product still generated meaningful revenue.

"If you're solving a real pain, people will adopt your solution even if it's incomplete. People are willing to go above and beyond to make your solution work for them, because the outcome is going to be insane for them. That, to me, was the signal," said Michel Tricot, co-founder of Airbyte.

According to Tricot, the speed was striking: one product hit $1 million in ARR within its first year, and another reached $1M ARR within just four months of launch. For a vertical AI startup, that's the core test — if your target industry will tolerate friction to get the outcome, you've found a real wedge. Incompleteness isn't the obstacle; indifference is.

How specific should a vertical AI startup's first customers be?

Extremely specific — your first customers should mirror the exact market you're targeting. Jay Madheswaran, founder of Eve (legal AI), shut down his old product lines to go all-in on a vertical AI product, and the conversion data validated the move.

"We noticed 40 percent conversion rates from cold outreach into demo requests. When we sent the 'we're shutting down the service product' email is when we got really strong signs of product-market fit, because people said no, don't take it away from me," said Jay Madheswaran, founder of Eve.

According to Madheswaran, Eve hit $1 million in ARR in its first quarter, then added another million two months later and another the month after that. He stressed careful qualification: with a barely-there product, you have to ensure early customers are "aligned with the larger market." For a vertical AI startup, a 40% cold-to-demo conversion rate is an extraordinary signal — and the "take it away and see who complains" test is one of the cleanest ways to confirm vertical fit.

Is a viral horizontal launch real product-market fit?

This is the most dangerous trap in AI right now. Rich White, founder of Fathom, learned that even 100,000 signups in a month doesn't necessarily mean you've found fit.

"We got 100,000 signups in the first month. It's over, we already won, right? But as we looked at our other metrics... The same problem I had doing user research calls was the exact same problem salespeople had. This isn't a vertical play — it's something everyone doing knowledge work on Zoom should have," said Rich White, founder of Fathom.

White's insight cuts both ways for vertical AI founders: Fathom started looking like a vertical (sales) tool but discovered the pain was horizontal across all knowledge work. According to White, he deliberately attacked an "800-pound gorilla" incumbent by coming in massively under their price — disruptive precisely because the incumbent couldn't drop revenue to match. The lesson: a vertical AI startup must honestly test whether its wedge is truly vertical or secretly horizontal, because the answer changes the entire go-to-market and pricing strategy. Vanity signups can mask the real question.

How do you charge for a vertical AI product early?

The pricing test for vertical AI is brutally simple: will someone pay at all? Dileep Thazhmon, founder of Jeeves, cut through the complexity founders create around early monetization.

"A lot of founders try to do this perfect business model. You're just a startup — can I pay you to use a product? If I can't pay you to use a product, there's no scenario where you can charge for it later. The core thing is, will someone use it, ideally for free, ideally then pay you," said Dileep Thazhmon, founder of Jeeves.

According to Thazhmon, Jeeves did $1 million in roughly six months and about $7 million a little past a year — and he did everything himself early on, from the UX to the logo, because founders "have to be 60% good at everything" to get a vertical product off the ground. He also normalized failure along the way: "if you don't feel like your startup has died three times, you're probably not running a startup." For vertical AI founders, the takeaway is to validate willingness to pay immediately rather than over-engineering a pricing model for a product nobody has committed to.

Do you need to be technical to build a vertical AI startup?

No — domain insight often matters more than a technical pedigree. On a PMF Show solo episode reflecting on founders who built their own way, Pablo Srugo pushed back on the myth that non-technical founders are doomed.

"In the early days I had this insecurity. We were two business founders, non-technical founders, and I felt like we were set for failure. When I say it out loud now, it's so stupid. There are so many successful business founders," said Pablo Srugo, host of the PMF Show, reflecting on his own startup Gymtrack.

According to Srugo, startup land is full of self-imposed limiting beliefs — "if I don't raise a big seed round, can I really be successful?" — that simply aren't true. For vertical AI specifically, deep understanding of an industry's workflow is frequently the harder, more valuable asset than the model itself, since modern tools make building an MVP more accessible than ever. The lesson: a vertical AI startup is often best built by someone who lives the industry's pain, technical background or not.

What's the vertical AI playbook these founders share?

Looking across the five stories, a repeatable vertical AI playbook emerges. It starts with finding an industry workflow painful enough that customers will tolerate an incomplete product — Michel Tricot's signal at Airbyte, where one product reached $1M ARR in four months. It continues with ruthless qualification of early customers so they represent the real market, the discipline Jay Madheswaran used at Eve while watching 40% of cold outreach convert to demo requests and stacking $1M ARR quarter after quarter.

The third move is validating willingness to pay immediately rather than over-engineering a model, as Dileep Thazhmon did at Jeeves on the way to roughly $7 million past its first year. The fourth is honesty about whether the wedge is truly vertical — Rich White's 100,000 Fathom signups looked like a vertical win but revealed a horizontal market across all knowledge work, which reshaped pricing and positioning entirely. And underpinning all of it is the recognition, voiced by Pablo Srugo, that domain insight often matters more than technical pedigree, since modern tools make building an MVP more accessible than ever. According to the founders on the PMF Show, the vertical AI startup that wins is the one that goes deepest on a single painful workflow, charges for it early, and stays honest about where the pain actually lives.

Key Takeaways

1. Acute pain beats a complete product. Airbyte found fit when customers adopted a bare-bones tool — one product hit $1M ARR in four months because the outcome was worth the friction.

2. Qualify first customers for market alignment. Eve saw 40% cold-to-demo conversion and used a "take it away and see who complains" test to confirm vertical fit.

3. Beware vanity signups. Fathom's 100,000 first-month signups didn't confirm fit — and revealed the wedge was horizontal, not vertical.

4. Test willingness to pay immediately. As Jeeves's founder put it, if someone won't pay (or use it free), there's no later scenario where you charge.

5. Expect to die a few times. Jeeves did ~$1M in six months and ~$7M past a year, but its founder says a real startup "dies three times" along the way.

6. Founders must be 60% good at everything early. Jeeves's founder did the UX, UI, and logo himself to get the vertical product off the ground.

7. You don't need to be technical. Deep domain insight into an industry's workflow is often the scarcer, more valuable asset for a vertical AI startup.

Listen to the Full Stories

This article draws from real founder interviews on the PMF Show, hosted by Pablo Srugo. For the complete vertical AI product-market fit stories, listen to the episodes with Rich White (Fathom), Michel Tricot (Airbyte), Dileep Thazhmon (Jeeves), Jay Madheswaran (Eve), and the PMF Show solo episode on founders who built their own way.

Last updated: June 2026