The Vertical AI Deal Book 2026: Valuations, Red Flags, and What Investors Are Actually Paying For

# The Vertical AI Deal Book 2026: Valuations, Red Flags, and What Investors Are Actually Paying For

Something unusual is happening in vertical AI: a small cohort of companies is publishing actual revenue figures alongside their valuations. Not ARR projections. Not pipeline. Revenue.

The Investors' Almanac has compiled what they're calling the Vertical AI Deal Book 2026 — a valuation-to-revenue analysis of six vertical AI companies that disclose numbers, alongside a framework for evaluating any vertical AI deal. The result is one of the more grounding documents to emerge from a funding environment that has been, to put it charitably, optimistic.

Here's what matters most.

The one company worth looking at first

Basis, an accounting AI company, sits at a $1.15 billion valuation. It deploys AI agents that prepare financial statements and file taxes. That's the entire business model: replace the work that accounting firms charge by the hour.

Basis is now part of a small group of vertical AI companies that have crossed the unicorn threshold while remaining operationally lean. No disclosed headcount explosion. No enterprise sales army. Just agents doing the work.

This matters for a specific reason. If you're tracking revenue per employee across AI startups, Basis is the type of company that breaks the conventional benchmark. When AI agents prepare the deliverable, the revenue-per-head ratio stops looking like a software company and starts looking like something that doesn't have a historical category.

What valuation-to-revenue actually tells you

The Deal Book's core contribution is simple: it lines up disclosed revenue against current valuation for six companies. The multiples vary significantly, and that variance is the story.

Vertical AI companies commanding the highest multiples tend to share a few structural traits:

• Deep workflow integration — they don't sit on top of existing tools, they replace a process entirely
• Contracted revenue — not usage-based speculation, but signed commitments
• Low model cost exposure — they've either fine-tuned to reduce inference costs or built pricing that absorbs volatility
• Liability clarity — they know who's responsible when the agent makes a mistake

Companies lacking even one of these trade at a meaningful discount. Lacking two or more signals a deal that has priced in hope rather than durability.

Ten questions that separate durable vertical AI from the rest

The Deal Book includes a ten-question diligence framework for vertical AI founders and investors. These aren't generic SaaS questions. They're specific to the structural risks that make vertical AI different.

The critical ones:

1. What is the model cost as a percentage of gross margin? Vertical AI companies running inference on every customer interaction have a cost structure that scales with revenue in ways traditional SaaS doesn't. If that number isn't capped or hedged, margin compression is a matter of when, not if.

2. Who holds liability when the agent is wrong? In accounting, legal, medical, and financial verticals, the output has real-world consequences. Companies that haven't resolved this contractually are carrying an undisclosed liability.

3. What percentage of revenue is contracted? Usage-based revenue in vertical AI can evaporate faster than it arrived. Contracted ARR is the signal that customers have committed to the workflow, not just the trial.

4. How deep is the integration? A product that sits inside an existing workflow and becomes hard to remove is worth significantly more than one that could be swapped out in an afternoon. Integration depth is moat measurement, not a technical detail.

5. What happens when the underlying model changes? Foundation model updates have broken vertical AI products before. Founders who don't have an answer to this question are building on ground they don't control.

The remaining five questions in the Deal Book cover customer concentration, data ownership, regulatory exposure by vertical, switching costs, and competitive response time from incumbents. All worth your time if you're evaluating or building in this space.

Why vertical AI is structurally different from horizontal AI

Horizontal AI tools — the writing assistants, the meeting summarizers, the general-purpose chatbots — compete on features and are bought by individuals. Vertical AI is bought by businesses to replace or augment a specific function. That changes everything about the sales motion, the pricing model, the competitive dynamics, and the valuation logic.

Vertical AI companies listed among the most credible AI-native businesses in 2026 tend to cluster around four sectors: legal, accounting, healthcare administration, and industrial operations. These are sectors where the output has measurable value, where the buyer is an organization rather than a person, and where the switching cost is high once the agent is embedded.

This is also why the revenue figures that exist tend to be more reliable than those in horizontal AI. Enterprise buyers don't sign contracts for tools they aren't using.

The solo founder angle

The Deal Book is aimed at investors, but the ten questions are equally useful as a founder stress test — particularly for solo founders building AI-native companies who are approaching their first institutional round.

The uncomfortable truth is that many vertical AI companies being pitched right now would fail three or four of those ten questions. Not because the product doesn't work, but because the founder hasn't resolved the structural issues that determine whether the business is durable at scale.

If you're a solo founder who has built an agentic product in a vertical, work through the Deal Book framework before you start fundraising conversations. It will either sharpen your pitch considerably or reveal the work you still need to do.

What the revenue disclosures actually signal

The six companies in the Deal Book that publish revenue figures are doing something most of their peers won't: they're giving buyers and investors a real number to price against. This is not altruism. It's a competitive signal.

Publishing revenue says the number is defensible. It says the company isn't hiding behind GMV, pipeline, or letter-of-intent counts. And in a market where many vertical AI companies are valued on narrative, a real revenue figure changes the conversation entirely.

Watch how many companies add disclosed revenue to their public profiles over the next 12 months. It will be a leading indicator of which part of the vertical AI market has matured past the story stage.

The number that matters most

For founders building in this space, one metric from the Deal Book framework deserves priority attention: contracted revenue as a percentage of total revenue. It's the number that tells you whether customers are buying the product or tolerating the trial.

Basis at $1.15 billion isn't valued on potential. It's valued on agents that do the work, in a vertical where the work is mandatory, recurring, and high-stakes. That's the template worth studying.

If you're building toward a similar profile — lean team, high revenue per head, agentic delivery — the revenue per employee metrics that define the AI startup generation are the benchmark worth tracking against your own numbers, not just the valuation multiples.

The Deal Book is free. The diligence framework is practical. Use both.

---

Is your company eligible? Submit to the leaderboard → onepersonunicorn.co/submit