Risk-based vs skill-based bundles.

A professional role bundle is held together by a binding constraint. For most professions that constraint can take one of two forms - and many fields run both modes in parallel for different sub-cuts of the work.

• Skill-based bundles are held together by scarcity. The rare person who can do all the components reliably. The market priced the whole package because separating it forced clients to manage multiple specialists. This is the default mode for most freelance professional work.
• Risk-based bundles are held together by institutional / legal / professional liability. Someone has to be sworn, certified, or accountable when the work drives a regulated outcome. The binder is the seat at the institution, not the rare skill - and the seat carries credentials, oaths, and recourse paths that the skill alone does not.

AI displaces the skill binder. As models get better at the components - translating, drafting, analyzing, diagnosing - the scarcity premium that bound them together weakens. Components scatter. The bundle disassembles.

AI does not displace the liability binder. AI has no legal personhood. It cannot be sued, sworn in, censured by a professional body, or named as the responsible party on a court filing. The seat that the bundle was built around remains a human seat - by structural fact, not by model limitation.

The asymmetry is permanent. Better models will keep displacing skill-based work. They will not touch risk-based positions because the binder is non-technical - it is conferred by institutions, maintained by credentials, and enforced by liability frameworks that require a human or institutional actor.

To predict how AI affects a given profession, separate the two modes and ask:

1. What share of the profession is skill-bound? Most commercial work in most professions - the bulk of legal drafting, medical chart-reading, accountancy, financial analysis. This is AI's target.
2. What share is risk-bound? The credentialed, institutional seats - court-certified interpreters, sworn translators, attorneys of record, attending physicians, audit partners, certified financial planners. This is AI's structural limit.
3. What's the regulatory trajectory? If regulation extends liability scope (the Joint Commission 2026 rule, the EU AI Act high-risk categories), risk-bound positions grow under AI pressure - not because AI created them but because the regulatory response to AI deployment did.