Aipedia: Accountants

AI in financial due diligence

Aipedia, by the Apex broking team · Last reviewed 2026-10-27

Accountants can use AI to review data rooms in financial due diligence, but they need to know what the tool did and did not read. A liability missed because coverage was incomplete can lead to a large claim from the buyer.

Part of: AI in Accountants

In short

On transactions, AI tools can read hundreds of contracts, extract key terms, analyse ledgers and flag unusual items far faster than a team of juniors. The risk is incomplete coverage: documents in formats the tool could not read, files added late to the data room, or issues the model was not asked to look for. The ethics Code requires accountants to understand the technology, recognise data limits (300.6 A2), check output is appropriate (R320.11) and explain limitations (R113.3). Deal claims can be large, so scope, reliance letters and PI limits all matter.

How accountants use AI in due diligence

Financial due diligence for a buyer or lender involves reviewing historical results, quality of earnings, working capital, net debt, tax exposures and key contracts. AI tools now assist with much of the reading: extracting terms from customer and supplier contracts, identifying change of control clauses, analysing ledger data for unusual journals, and summarising board minutes. Generative AI may draft sections of the report.

Deal timetables are tight, and AI lets a smaller team cover more documents. That makes it more important to know exactly which documents were covered and how. Team members who would once have read every contract may now review a summary table, so the firm's quality control has to move to checking the tool's coverage, its extraction accuracy on a sample, and the questions it was set.

The duties that apply

113.1 A3 of the IESBA technology revisions: competence includes understanding the technology you rely on.
300.6 A2: the data might not be sufficient for effective use of the technology, for example scanned documents the tool cannot read accurately.
R320.11: decide whether output is appropriate for the purpose. A clause extraction tool tuned for one jurisdiction may miss terms drafted in another style.
R113.3: explain the limitations of the work, which in due diligence means the scope and coverage of the review.
Confidentiality: data-room material is highly confidential, and the CCAB draft Statement recommends enterprise-grade tools and specific consent before data goes into AI tools.

What can go wrong

Illustrative scenario: a firm uses an AI review tool on a target's data room. A batch of supplier contracts was uploaded as image scans late in the process, and the tool's extraction from them was poor. One contract contained a significant termination payment triggered by a change of control. The report did not mention it. After completion the supplier enforced the clause, and the buyer claimed against the firm for the cost.

Similar problems arise when the tool is asked to find certain clause types and the important issue is something else, or when AI-generated summaries of board minutes leave out a dispute that was mentioned in passing.

The report is often relied on by more than one party: the buyer, its lenders and sometimes investors. Each may say it relied on the firm's work, which can multiply the potential exposure from a single oversight.

How PI usually responds

A buyer claim for a missed liability usually falls within an accountant's civil liability PI, subject to the terms, excess and limit. The issues are scope, reliance and causation: what the firm agreed to review, who could rely on the report, and whether the buyer would have paid less or walked away. Reliance letters, liability caps and clear scope wording are the main protections; see engagement letters and liability caps and Caparo for accountants. Deal claims can exceed a firm's minimum limit, so review how much PI cover you need and whether excess layer cover makes sense.

Practical controls

Reconcile the data-room index with what the tool actually processed, including late uploads.
Have a person review documents the tool could not read well.
State in the report which documents were reviewed by AI and which by people.
Agree reliance and caps in writing before the report is released.
Use the same scepticism as with anomaly detection and keep valuation work aligned with the findings.
Explain AI use in the engagement letter.

How Apex helps

Apex arranges PI for accountants who carry out transaction work, and a broker can review limits and any excess layers against deal sizes. See the accountancy AI overview.

Sources

Frequently asked

Is AI document review acceptable in financial due diligence?

Yes, if the firm understands the tool, checks its coverage and output, and explains the scope of the review. The Code requires competence with technology and a judgement on whether output is appropriate before relying on it.

How can firms limit exposure on due diligence reports?

Agree a clear scope, set out who may rely on the report in reliance letters, and agree a liability cap where appropriate. State the limitations of the review, including how AI tools were used and which documents they covered.

Can one due diligence claim exceed our PI limit?

Yes. Transaction claims can be large compared with the minimum limits set by ICAEW, ACCA or AAT. Firms doing deal work often review whether their limit fits the transactions they advise on and consider excess layers.

Should we tell the client which documents were reviewed by AI?

Yes. R113.3 requires you to explain the limitations of the work, and the CCAB draft Statement favours openness about AI use. A clear statement of coverage also helps if the scope of the review is later disputed.

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.