AI and the insurance operating model: costs and claims before underwriting?
Where will AI change insurers' economics first?
- Horizon: 12 to 36 months
- Direction: Uncertain
- As of
- Demonstration
An outlook is a view held with stated confidence. It is not a forecast, a prediction of financial results, or a recommendation.
The view
The working thesis is that the material effect of AI on insurance will show first in expense ratios and claims-handling costs, and only later in underwriting selection and pricing. It is a thesis to test, not a finding: current disclosures cannot yet confirm it, and the one related transaction recorded here, a group buying an insurer it describes as having a digital underwriting platform, does not distinguish between the two: ERGO's releases mention AI only in a background description of NEXT, as used to simplify purchasing, and disclose no effect on costs, claims or underwriting results.
What we expect
- 01
Measurable effects appear first in expense ratios and claims handling.
Low confidence - 02
Effects on underwriting selection and pricing follow later and are harder to attribute.
Low confidence
Evidence
Transaction · ERGO (Munich Re) · 20 March 2025
ERGO gains access to the U.S. market with the full acquisition of NEXT Insurance
SupportsERGO's description of NEXT's fully digital underwriting platform. ERGO's releases mention AI only in a background description of NEXT, as used to simplify purchasing, and disclose no effect on costs, claims or underwriting results, so they do not distinguish the thesis from its alternative.
Implications
Capital implication
If the thesis holds, the first financial evidence would be in expense ratios and claims-handling costs, before risk selection shows in loss ratios.
Strategic implication
Groups would compete first on operating cost and claims efficiency; underwriting advantage from AI would take longer to show.
Risks to the view
- Disclosures continue not to separate AI-driven gains from other efficiency programmes, or keep using AI terminology without measurable change in reported ratios, so the thesis cannot be tested.
- Underwriting and pricing effects appear earlier than expected, for example in personal lines with dense data.