Five signals identify which services are ripe for agent replacement

Not every service moves at the same speed. Some are being eaten this year. Some will be eaten in 2028. Some categories may resist replacement for the foreseeable future where regulation, liability, or human-in-the-loop requirements are binding.

The diagnostic is five signals. A service that hits four of them with no severe regulatory blocker is being prototyped in a YC batch right now — probably more than once.

Signal	Why it matters
The work is repeatable	Same shape, different inputs. Agents excel at pattern execution, not improvisation.
The work is already outsourced	Customer has accepted that someone else owns execution. Swapping one outside vendor for another is a small step.
There's a clear right answer (or a tight band)	Tax returns either pass an audit or they don't. Strategy recs are fuzzier. The first gets eaten faster.
Customer measures success by outcome, not effort	Nobody cares how many hours their accountant worked. They care the return was filed correctly.
Margins are high enough to attract a vertical agent	A $50/hr service is hard to disrupt; labor is already cheap. A $500/hr service has room for software-priced competition to look magical.

The absence of a regulatory blocker matters as much as the presence of the signals. Insurance underwriting and healthcare diagnosis have several of these signals but face constraints that slow or block replacement regardless of model quality.

Source claim: Five diagnostic signals — repeatable, already outsourced, clear right answer, outcome-measured, high-margin — identify which services are being prototyped as vertical agents right now, and regulatory constraints are the primary override.