LO
Insurance claims handler — how we know
The page itself gives the judgements. This one gives what they rest on: which technologies bear on the work, how the estimate moved since language models reached the public, and the method behind both.
Which technologies matter here#
Four separate signals. They are deliberately not added together — a job exposed to two technologies is not twice as exposed.
How it got here#
The index is not a static number. This is where it would have sat at each capability checkpoint since ChatGPT — reconstructed, and labelled as such.
—— this stretch contains a verified event- - - no event in this stretch — reconstruction only0 = no task exposed, 100 = every task exposed
● 2 verified events for this occupation, plotted at the date it happened — the parts of the line near a marker are anchored to something checkable.
A high start that predates generative AI: rules engines and straight-through processing have been eating simple personal-lines claims since the 2000s. The 2023-2025 climb is the intake end — reading a free-text account of what happened, matching it to a policy and settling it — which had resisted rule engines because it needed language. One insurer's own annual report puts 98% on that step. It flattens because the remaining work is not processing: the claim whose story does not add up, and the person who has to answer for a decline. Read the level as a warning about the routine half rather than about the occupation, and note what the curve cannot see — the number of people needed per thousand claims, which is the number a business case is actually built on.
A flat line is not a forecast of safety. It says which tasks automation has reached so far — the occupations that moved least here are the ones where the constraint is physical or regulatory, and both of those can change.
Method and sources#
- Assessment date
- 2026-09-13
- Basis of the task judgements
- 3 evidence-backed · 3 platform inference · 0 not enough evidence
- Verified events
- 2