Three service layers predict which work AI eats first
Every service business has three layers. Two are getting absorbed into Service as a Software companies right now, at very different speeds. One is not.
Layer 1 — production work — is being eaten first, fastest, and most visibly. Filing the tax return. Drafting the contract. Posting the invoice. This is hours-for-dollars execution: repeatable, measurable, and already showing up in agent adoption data for software engineering. The same shape is starting to play out in other production-heavy categories.
Layer 2 — pattern application — is next. Translating a known problem into a working answer using a playbook the industry has refined for decades. The pace here is highly domain-specific. Verticals with clean ground truth (tax, payroll, certain compliance filings) move fast. Verticals with binding regulatory or trust constraints (healthcare diagnoses, legal advice, insurance underwriting) move slowly, sometimes much slower.
Layer 3 — strategic direction — is not being absorbed in any meaningful way yet. Picking the right problem. Naming what the customer actually needs. Deciding what to refuse. This layer is getting more valuable, because the cost of executing Layers 1 and 2 is collapsing toward the marginal cost of inference. A business that sells Layer 3 plus packaged 1+2 outcomes is positioned to be the agent. A business that sells Layers 1 and 2 by the hour is competing against agents that ship the same work for an order of magnitude less.
| Layer | What it is | AI absorption pace |
|---|---|---|
| 1 — Production work | Repeatable execution: file, draft, post, send | Fast — happening now |
| 2 — Pattern application | Known-problem-to-known-answer using domain playbook | Domain-dependent; regulated verticals slow |
| 3 — Strategic direction | Naming the right problem, framing the engagement | Not absorbing; becoming more valuable |
Source claim: The three-layer model of any service predicts absorption order: production work goes first, pattern application goes next at domain-dependent speed, and strategic direction is not being replaced — it is appreciating in value.