Aipedia: Claims management companies

AI review of agreements and policy documents

Aipedia, by the Apex broking team · Last reviewed 2026-10-30

CMCs can use AI to read credit agreements and policy documents for mis-selling evidence, but the firm is responsible if the analysis is wrong. Errors repeat across every file, turning one fault into many claims.

Part of: AI in Claims management companies

In short

AI document review extracts rates, commission, premiums and dates from agreements and statements, and some tools calculate redress. The two main errors are false positives, which lead to weak complaints and lost Ombudsman fees, and missed grounds, which can cost consumers their claims. The Consumer Duty and FCA accountability for outsourced AI put responsibility on the firm. A consumer who loses out through negligent analysis may claim against the firm. PI may respond if held, subject to the wording and aggregation terms, but re-review costs are usually not covered.

How CMCs use AI to review documents

Finding evidence of mis-selling used to mean a person reading every credit agreement, policy schedule and statement. AI document review tools now do the first pass: they classify documents, extract interest rates, commission disclosures, premiums, dates and balances, and flag possible grounds for complaint. Some go further and calculate the redress a consumer might be owed. The tools are fast and consistent, but they can be consistently wrong.

The duties that apply

Consumer Duty: acting in good faith (PRIN 2A.2.1R) and avoiding foreseeable harm (PRIN 2A.2.8R), in PRIN 2A.2, apply to the analysis the firm relies on and passes to the consumer.
Accountability for tools: the FCA's good-practice page says firms deploying AI "remain responsible for consumer outcomes" and can outsource AI services but remain responsible for them.
Records: SYSC 9.1.1R supports keeping the extracted data, the source and the reviewer's checks.
Data protection: the documents contain personal financial data, so the ICO's AI guidance and UK GDPR apply to the tool and the vendor.

False positives and missed grounds

Two errors matter. A false positive flags a ground that is not there, for example reading a disclosed commission as undisclosed. The firm then pursues a weak complaint, pays the Ombudsman case fee if it refers and loses, and leaves the consumer disappointed. A missed ground is worse for the consumer: the tool fails to spot a feature that supports a claim, and the firm either rejects the case or argues it on a weaker basis. Both errors tend to repeat across every file with the same document layout, so a single fault can affect a whole book of claims.

The Ombudsman has already complained about inaccurate AI material from professional representatives, as covered in AI-written complaints to the Ombudsman.

What can go wrong

Illustrative scenario. A CMC's AI tool extracts APRs from scanned credit agreements. For one lender's older template it reads the monthly rate as the annual rate. The redress calculation in each complaint letter is overstated, the lender rejects the figures, and some consumers accept a lower offer believing the firm's numbers were inflated. Other agreements from the same lender were scanned at low resolution and the tool missed a commission clause altogether, so those consumers were told they had no claim.

Redress figures deserve special care. A number in a complaint letter looks authoritative to the consumer, who may use it to judge whether an offer is fair. If the figure is wrong in either direction, the consumer may reject a reasonable offer or accept a poor one, and the firm's credibility with the lender and the Ombudsman suffers on every later case.

How PI usually responds

A consumer who can show the firm's negligent analysis caused them to lose or under-recover a claim may bring a negligence claim, and PI, where held, may respond subject to the wording. Because the same fault runs through many files, check how your policy aggregates related claims, which affects both the limit and the number of excesses. Remediation costs, such as re-reviewing a book of files, are not usually covered unless the policy has a specific mitigation extension. Our guides to whether PI covers AI mistakes and claims firm PI explain more.

Practical controls

Validate the tool on each lender's document templates before relying on it.
Have a person check every redress figure before it goes in a letter.
Review a sample of "no claim" outcomes as well as positive flags.
Flag low-quality scans for manual review.
Record the tool version, source document and reviewer for each file.
Feed findings into eligibility screening and DSAR handling controls.

How Apex helps

Apex arranges PI for claims management companies and can review how your policy handles aggregated errors from a single AI tool. Read the overview in AI in claims management.

Sources

Frequently asked

Can a CMC rely on AI to find mis-selling evidence?

It can use AI to speed up review, but the firm remains responsible for the analysis. Validate the tool on real document templates, check figures before they are used and review negative outcomes as well as positive flags.

What happens if an AI tool misses a valid claim?

The consumer may be told they have no claim or have it argued on a weaker basis. If they can show the firm was negligent and they lost out, they may bring a claim against the firm, which PI may cover if held, subject to the wording.

Should redress figures produced by AI be shown to consumers?

Only after a person has checked them against the source documents and the method used. Consumers use these figures to judge offers, so an unchecked number can cause real harm. Label any figure as an estimate where it is one, and keep the calculation on file.

Does PI cover the cost of re-reviewing files after an AI error?

Not usually. PI generally responds to third-party claims and defence costs. The firm's own costs of re-reviewing files are commonly not covered unless the policy has a specific mitigation or loss prevention extension.

Check your PI covers how you use AI

Tell us about your firm and the work you do. A broker will review how your current PI treats AI-assisted work, or arrange new cover. No fee for a quote. Or call 0117 325 0027.

Start a PI quote Request a callback

Apex Insurance Brokers Limited is authorised and regulated by the Financial Conduct Authority. Registered in England and Wales, company number 07014570. This page is general information about how AI is used in the profession and how insurance may respond. It is not legal, regulatory or insurance advice on your circumstances, and it does not guarantee that cover will be available or on what terms.