01What happens.
A promotion page answers the question a customer asks in plain terms, while the policy document holds the limits, conditions and exclusions that govern the claim. AI serves the page. The customer receives the headline cover, and the claim is settled on the policy AI did not present in its answer.
02Why it matters.
For the people who run claims and customer operations, the customer who needs the policy now asks AI what it says before they call. If the answer comes from a page marketing published two seasons ago, the cover the customer believes they have is the headline cover, without the limits that apply to the claim, and the number they call for help no longer reaches the assistance service.
The complaint that follows is about the claim, and the cause is a page nobody has looked at since the campaign closed.
03The case.
The insurer is anonymised, and the case is an observed AAAnow finding published in the AI Readiness paper.
An insurer ran a holiday promotion for its travel cover. The promotion page set out the cover in the plain terms a promotion uses: medical expenses up to a figure, cancellation up to a figure, baggage up to a figure, and an emergency number to call. The policy document, which governed the contract, set out the cover in full, with the limits, conditions and exclusions that the promotion did not carry. The promotion was a headline account of the cover, and it had never been more than that.
The promotion ended and the page stayed online.
Later, the insurer changed its emergency assistance arrangements and moved to a single global freephone number. The policy documents were updated with the new number. The promotion page was not, because nobody owned it once the campaign had closed, and it still carried the old number.
Customers preparing to travel, and customers already abroad, asked AI what their policy covered and whom to call. The answer came from the promotion page. It returned the cover levels as the promotion had stated them, without the limits and conditions the contract applied, and it returned the emergency number as it had been, which no longer reached the assistance service.
A customer needing assistance abroad called the old number and reached nobody who could help. A customer making a claim found that the headline figure AI had quoted was subject to a limit in the policy that AI had not mentioned. Each had acted on the insurer's own published words, and each had received the wrong ones from a machine that chose the readable page over the filed document.
The insurer heard about both through complaints, after the event.
Marketing owned the page and closed the campaign. Product owned the policy and updated it. Assistance owned the number and changed it. No team had a view of what AI would weight when a customer asked, and the readable page won each time.
04Exposure.
The insurer answers for the cover it describes to customers, on the policy and on the page alike. A customer shown the headline cover without its limits, and a wrong number, at the point of need has a complaint the insurer cannot answer with the contract alone, and the regulator's interest in consumer outcomes reaches this point: what the customer was told, by material the insurer publishe d, when it mattered.
Complaints, claims disputed on figures the customer had been quoted, and customers who could not reach assistance when they needed it all follow, and beyond them the insurer loses trust in its word at the moment the customer is relying on it, which is the moment the policy was bought to cover.
05What AI Readiness changes.
Inside the insurer, the policy was the policy and the promotion was over. Nobody was placed to see that the promotion page had become the policy in the machine's reading, and that is the view AAAnow supplies: the insurer's digital landscape as AI reads it, with the pages and documents AI is drawing on for customers' questions evidenced, where those pages fall short of the policy, and which contact details are being served.
The insurer can then decide, product by product, which document stands, retire the pages that contradict it, and check the position before the next season's promotion goes live.
Your team makes the changes; AAAnow provides the assessment, the evidence, the guidance and the review.
06Related use cases.
Regulator's questioned governance where the compliance register stopped
The inquiry that can follow a customer complaint.
Residents get different answers to the same question, and the cost lands on the council
The same contradiction across an estate, in a public body.
Announcements buried in the page furniture
A change published and never heard.
07References.
- AAAnow, AI Readiness paper, Jun. 2026: documented use case, anonymised.
- Gallagher, AI adoption and risk benchmarking 2026. www.ajg.com/news-and-insights/features/ai-adoption-and-risk-benchmarking-2026ajg.comWe do not enable external links - please copy and paste, you are then certain as to the link being opened
Next step
Talk it through, in your terms.
The briefing sets this case in your organisation’s terms, and the discussion follows your questions.
Observed, anonymised. No firm is identified.