Vertical AI agents quietly crossed $100M ARR while we argued about models

These aren't pilots anymore. Sierra reports $150M+ ARR within eight quarters of launch (on $950M raised), Intercom's Fin crossed $100M+ ARR resolving ~2M issues a week, and EliseAI sits around $100M ARR on $250M raised. Three companies selling completed work instead of seats, all at genuine commercial scale.

The structural shift underneath: 47% of the top 500 U.S. enterprises moved at least one business process from traditional SaaS to a vertical AI agent across 2024-2025, up from just 11% in 2023. CIOs are managing a forced re-platforming they didn't ask for. And the incumbents are disclosing their own agent revenue now, which tells you it's material: ServiceNow claims ~$600M of agentic revenue, Salesforce ~$169M growing 800% YoY, Workday $400M of its $8.8B ARR. Still single-digit percentages of total ARR, so the incumbents are monetizing AI as an add-on, not yet seeing it cannibalize the core. The displacement is real but the transition is early. Both things are true.

The wedge is consistent: deep domain integration plus outcome-aligned pricing. A single branded agent replacing a multi-seat support or CRM deployment. You see the same logic in Salesforce shipping Agentic Advisor defensively because AI notetakers Jump and Zocks are capturing the advisor-client relationship and bypassing the CRM entirely. When the AI lives in the meeting, the system-of-record becomes a passive database.

If you're building, the takeaway is sharp. Per-seat pricing in a category an agent can do end-to-end is a depreciating asset. The companies winning sell the outcome (resolved ticket, leased apartment, closed deal) and price against the labor they replace, not the login count. Pick a narrow vertical, integrate deeper than an incumbent reasonably can, and charge for results.