After Harvey, vertical AI’s next $10B winner might be in agriculture

The article highlights rapid AI startup valuation growth: Harvey reportedly surpassed an $11B valuation, Sierra is valued at over $15B, and Abridge has built a multi-billion-dollar business converting clinical conversations into structured medical records. It frames these as fast-moving, high-growth milestones, suggesting continued investor appetite for applied AI products.

Analysis

The market is still underpricing how quickly AI agents move from demo-quality to workflow control in regulated, high-friction verticals. The immediate winners are not the private app names themselves but the public companies that own proprietary content, system-of-record distribution, or enterprise trust: legal information/workflow, EHR/healthcare software, and customer-service software can all monetize AI as an attach-rate and retention tool rather than a standalone product. The obvious losers are labor-intensive service providers and BPOs where AI can remove 10-20% of seat demand before revenue rolls over, creating a lag where margins compress faster than reported growth.

The 1-3 month catalyst is not usage hype; it is earnings commentary on net retention, gross margin, and implementation backlog. If incumbents show AI driving faster case resolution, shorter handle times, or higher conversion into premium tiers, the multiple expansion can be real even without large revenue re-acceleration. Over 6-18 months, the bigger effect is budget reallocation: spend migrates from headcount-heavy outsourced work toward software with proprietary data and workflow embedded, while generic point solutions get bundled or acquired.

The contrarian view is that these valuations may reflect scarcity, not durability. In legal and healthcare especially, liability, data rights, and integration depth are the moat; many standalone agents will struggle to own distribution once incumbents ship comparable features. That means the best public expression may be a long/short on incumbents versus labor proxies, not a simple long AI beta trade. Falsifiers: weak AI attach in the next two earnings cycles, rising churn at the incumbents, or any regulatory push that slows automation in healthcare/legal workflows.

Key Decisions for Investors

• Long RELX / short CNXC or TTEC for 1-3 months: express the view that AI deflects low-complexity service volume faster than it displaces proprietary workflow/data franchises; target a 5-10% relative move if commentary turns on margin mix.
• Long NICE on pullbacks over 6-12 months: if AI agents reduce contact-center handle time without impairing outcomes, NICE can monetize via higher software spend while BPOs lose labor leverage; invalidate if customer churn or seat contraction appears in the next two quarters.
• Long Thomson Reuters as a slower, higher-quality beneficiary of legal AI adoption: the upside is less about headline growth and more about pricing power and retention; best entry is after any market disappointment that overreacts to near-term implementation noise.
• Watch VEEV for a 6-18 month rerating, not a quick trade: if AI documentation is real, the value pools to system-of-record vendors with regulatory trust and workflow lock-in; this is attractive only if management proves AI lifts sales efficiency and expansion revenue.
• Avoid buying pure-play AI application names solely on private-market marks: use them as sentiment indicators, not as proof of public-market upside; the trade fails if incumbents absorb the use case faster than standalone companies can scale.