'AI for everyone' is the new 'mobile-first'. Once it's a category, it's no longer a wedge.
W25 made it explicit: roughly two-thirds of AI-as-agent companies in the batch target one industry or one role, not a horizontal productivity slice.
Where the vertical bets are concentrated
- Legal: contract review, paralegal workflows, immigration filings.
- Healthcare ops: prior auth, claims appeals, RCM workflows.
- Customer support: tier-1 ticket resolution with auth-scoped tool use.
- Sales ops: pipeline hygiene, account research, outbound personalisation.
- Field operations: dispatch, scheduling, work-order closeout.
Why horizontal died as a wedge
Horizontal copilots compete with every model lab's free tier and every incumbent's bundled feature. The CAC math stopped working in 2024 and YC noticed.
Vertical agents win on three things horizontal can't touch: domain-specific evals, regulated-workflow trust, and integrations into systems of record nobody else has bothered to wire up.
What this means for your application
- Name the industry in the first sentence of your application. 'AI for sales teams' is too broad. 'AI SDR for outbound to dental practices' lands.
- Show one workflow end-to-end, not five workflows partway.
- If you can't name the integration that makes you defensible, you don't have a wedge yet.
Key takeaways
- Vertical-agent companies outnumber horizontal copilots ~2:1 in W25.
- Horizontal wedges lose to free tiers and bundled features.
- Specificity in the application — industry + role + workflow — is the new signal.