Vertical AI is the category that produces the most investment theses and, proportionally, the fewest durable businesses. Here’s an honest scorecard of what’s actually working, and where the bodies are buried.
Finance: Real product, real revenue, real competition
Legal AI is the clearest success story in vertical AI. Harvey, EvenUp, and several smaller players have actual paying customers, measurable productivity gains, and renewal rates above 85%. The workflows are well-defined. The data is relatively clean. The output (legal documents, case analysis, contract review) is checkable by a human expert in a reasonable time. These conditions produced a real market.
Financial services is following the same playbook two years behind. The Anthropic announcement this week is the clearest signal: specialized models, domain-specific data partnerships, workflow integrations for the specific tasks banks care about (credit analysis, regulatory filing, earnings call prep). The first-mover advantage in finance is up for grabs in 2026.
Healthcare: Revenue cycle yes, clinical workflows no
Healthcare AI is bifurcated. Revenue cycle management — billing, coding, prior authorization — has seen genuine AI adoption. The tasks are repetitive, the rules are well-defined, and the ROI is immediate. Vendors in this space are generating real revenue.
Clinical AI is a different story. The liability exposure is real. The procurement cycle is 18–24 months minimum. The physician workflow integration challenge is genuinely hard. And the patient safety implications mean that every deployment needs human oversight infrastructure that most vendors haven’t built. Several high-profile clinical AI companies raised on clinical promise and are quietly pivoting to revenue cycle. Watch that space over the next six months.
Supply chain: Infrastructure plays are winning
Supply chain AI is having a moment, but the winners are infrastructure plays, not application plays. The companies generating the most durable revenue are the ones helping large enterprises clean and normalize their supply chain data, not the ones trying to predict demand or optimize routing on top of dirty data. Data infrastructure before AI features. That’s the sequence that’s working.
Where consolidation is coming
Legal AI is mature enough that the first wave of consolidation is imminent. Expect 2–3 acquisitions of Series B legal AI companies in the next 12 months — either by law firm management software incumbents or by the large legal publishers (Thomson Reuters, LexisNexis) who are now clearly in acquisition mode. The acqui-hire premium on legal AI talent is already visible in comp packages.
Finance AI is 18 months behind legal on the consolidation curve. Build for it now.