# AI Eats Services Not Software > Coatue put a number on what the best vertical AI companies already know: software is a $0.2T market. Services — BPOs, agencies, professional services, staffing — are $5.5T. That's a 25x delta. The shift is from selling tools to selling work. The competition isn't Salesforce or Workday. It's the $80K/year agency your client replaces with a $2K/month AI agent. --- ## Three Emerging Pricing Models Three business models are converging around this insight, each one pricing ***output*** rather than *seats*: - **Enterprise model** — base recurring license for the AI platform, plus a separate fee for the human deployment team - **Outcome model** — standard tier for raw AI output, premium tier where clients pay per guaranteed resolution handled by humans - **Fractional employee model** — the entire package priced as a fractional employee salary; AI handles bulk workload at near-zero marginal cost, humans manage edge cases The structural shift is the same across all three: you're billing for a result, not access to a tool. --- ## How Deep the Penetration Actually Is Anthropic's "observed exposure" metric separates theoretical AI capability from real-world coverage today: | Occupation | Theoretical | Observed | |---|---|---| | Computer & math | 94% | 33% | | Office & admin | 90% | 14% | | Business & finance | 87% | 18% | | Computer programmers | ~90% | 75% | | Customer service reps | ~80% | 70% | ![[Pasted image 20260329150201.png]] The frontier is jagged — AI is shockingly good at some tasks and poor at adjacent ones, and the boundary is unpredictable. But it moves fast. The gap between theoretical and observed is the investment opportunity. One leading indicator: hiring for 22–25 year olds in exposed occupations has already slowed 14%. Entry points to the labour market are narrowing before unemployment data shows any signal. See: [[Observed Exposure — Anthropic's AI Penetration Metric]] --- ## The Gap of Imagination Alvaro Higes (Luzia, 65M users) names the adoption problem precisely: even when AI can do something, most people can't picture what *good* looks like, so they can't ask for it. ![[92c4e814-870d-401b-80ac-3ebfa0cfc3af_3444x1924.webp]] Most users are still interacting with today's models the way they would have in November 2022 — ask it to tidy a calendar, maybe book a flight if they've built up enough trust. But a great EA doesn't just manage schedules. It frames decisions before you have to make them, filters ruthlessly what deserves your attention, and closes every loop. AI delivers roughly 70% of that today. Most users are getting 0%. The gap between 0% and 70% is not a model problem. It's an imagination problem. Two product failures drive it: 1. **Builders default to chat.** Chatbots look identical to each other and optimise for demos. The real wins come from AI embedded into workflows — boring, reliable, frictionless. 2. **Nobody closes the imagination gap.** If users can't picture the ceiling, they settle for the floor they stumbled onto. Christensen's rule holds: to change a default, your experience must be 10x better — not "AI-powered" better, but actually better in a way users feel immediately. See: [[The Gap of Imagination]] --- ## So What? The business model opportunity is in services, not software. The moat question shifts accordingly: can you build the AI-plus-human delivery model that actually guarantees outcomes at a price point that undercuts human-only alternatives? Companies that price per output *and* close the imagination gap for their buyers win on both ends of the problem. --- **Related:** [[AI Disruption Risk Is Not Uniform - Thoma Bravo]] | [[AI era Defensibility]] | [[AI-first GTM]] | [[Observed Exposure — Anthropic's AI Penetration Metric]] | [[The Gap of Imagination]] Tags: #investing #AIstrategy #businessmodel #labor #verticalAI