AI in insurance, checked dailyThursday 17 September 2026
The Renewal Report
News, findings and tests. Every item with its source, its evidence and what it means for a book of business.For agencies, MGAs and carriers
Finding 005Open, revised when AI-attributed claim data appears

The agency E&O loss data that already exists, and what AI does to it.

Question
Whether there is any data on agency E&O claims caused by AI, and if not, what the existing cause data implies
Short answer
No AI-attributed agency E&O data exists in anything we could find. The 2017–2022 cause ranking does, and an unreviewed AI summary sits inside its top two causes.
Basis
Big “I” Professional Liability / Swiss Re Corporate Solutions program data as reported in Independent Agent, 8 January 2024; Utica National, June 2024; defense counsel, 9 September 2026
Tier
Editorial interpretation — the ranking is program data with sample size not disclosed; the mapping is ours
Scope reviewed
One carrier program's cause ranking; not a market-wide loss study
Reviewed
16 September 2026

The finding

Nobody has published an AI-caused agency E&O claim count, frequency or severity. That is a search result, not proof the number is zero. What has been published is the cause ranking for a large agents E&O program over six years, and the ways an AI-assisted proposal fails map onto its top entries without any stretching.

The Big “I” Professional Liability program, underwritten through Swiss Re Corporate Solutions, reported its leading claim causes for 2017–2022 as, in order: failure to procure coverage; failure to adequately explain or disclose policy provisions; inaccurate information provided to the carrier; failure to recommend the correct coverage type; failure to recommend adequate value or limits. The sample size behind the ranking is not disclosed.

The mapping

Program cause ranking, 2017–2022, against the AI failure mode that produces it — the right column is ours
RankCause, as reportedHow an AI-assisted proposal produces itPre-send check that catches it
1Failure to procure coverageA coverage part in the quote packet silently omitted from the summary; the reader never knows it was thereSilent-omission check — walk the packet, not the output
2Failure to explain or disclose policy provisionsA twelve-month actual-loss-sustained limit becomes “Included”; an exclusion becomes a headingCoverage and exclusion comparison against form wording, edition recorded
3Inaccurate information provided to the carrierA value inferred by the model and presented as quoted; a location count that does not match the ACORDMissing-data check — “not provided,” never inferred
4Failure to recommend the correct coverage typeA comparison drafted from the wrong form family, or from a specimen rather than the quoted formCitation check against the actual document
5Failure to recommend adequate limitsA renewal diff that does not flag a reduced limit or a new sublimitVersion comparison against the expiring program

The ranking measures nothing about AI. It does not establish that AI raises claim frequency, and it cannot be used to quantify expected loss. The mapping is editorial: it says where an unreviewed output would land if it caused a claim, not that it will.

What the carrier and counsel say

Agency management must develop a Generative AI Use Policy that provides clear guidance on how GenAI tools should and should not be used at the agency.

Utica National E&O Risk Management Newsletter, Vol. II Issue 6, June 2024 — carrier loss-prevention guidance; the page disclaims legal advice and coverage

Defense counsel's list for the file is five items: the client's request, the information available at the time, the AI output, what was verified or changed, and what went to the client. Neither source is a policy condition or a statute, and neither reports an AI-specific loss. Both are the closest thing to loss-prevention data the market has.

What to do with this

Run the eight-test pre-send checklist on every AI-assisted proposal, and add one line to the file: “Drafted with AI assistance, reviewed and adopted by [name], [date].” The two together put a name and a date between the model and the client, which is the thing the top two causes on the ranking are missing.

Sources: Olivia Overman, “How the hard market and talent shortages compound agency E&O exposures,” Independent Agent, 8 January 2024 (Big “I” Professional Liability / Swiss Re Corporate Solutions program, 2017–2022, sample size not stated); Utica National, June 2024; Ventura & Gittleman, Independent Agent, 9 September 2026. Not found: any insurer-published AI-attributed producer E&O counts, any enforcement action against a producer for AI-drafted advice, any agents E&O form with an AI exclusion or affirmative grant.