Agent liability underwriter
When an agent books the slot or signs the order, someone has to price the mistake. No market for that exists yet.
- Horizon
- not yet a title
- Protection from automation
- Partly protected · our reading
This is the most speculative profile on the site, and the only one with no claim attached — because no public data describes a market that does not exist. What exists is the question. Agents are starting to take real actions in the world: booking the appointment, sending the payment link, confirming the order McKinsey 2025. Every deployment currently answers “who pays when the agent is wrong?” informally — a term of service, a human in the loop, a hope.
Industries have a way of answering that question formally. Insurance exists because unfamiliar risk eventually gets priced. An underwriter for agent actions would read an operator’s logs, guardrails and rollback design the way a car insurer reads a driving record, and put a number on the residual risk of delegation.
Why include something this speculative? Because the four tests read unusually well: total accountability, a relationship with the insured, an indivisible seat. If delegation keeps growing, the scarcity is not technical — it is people willing to stake a professional judgement on a machine’s behaviour. That scarcity is the job.
- Delegating real actions — payments, bookings, orders — to agents creates a liability question every deployment currently answers informally
- Insurance has a track record of turning new technologies into priced products
- Contract law may absorb the question by pinning liability on existing roles
- Agents that log everything may make the risk priceable enough that no human underwriter is needed