Aipedia: Actuaries

AI in peer review under APS X2

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

Actuaries can use AI to support Work Review and Independent Peer Review, but the IFoA advises caution and the reviewer must still exercise their own judgement under APS X2. A review that relies on AI without real challenge offers little protection if the work is later criticised.

Part of: AI in Actuaries

In short

APS X2 v1.1, in force from 30 January 2026, requires IFoA members to consider whether Work Review, and in particular Independent Peer Review, is appropriate and proportionate, weighing complexity, significance, likelihood of error and the judgement involved. It does not mention AI, but AI-heavy work is often more complex and more error-prone, which points towards more review. The IFoA data science and AI guide asks members to consider peer review and to be cautious about using AI within reviews. A well-documented human review is strong evidence of care in any later negligence claim; a superficial one is not.

Where AI meets review

AI enters the review process in two ways. First, the work being reviewed may use AI: machine learning models, generated code and AI-drafted reports. Second, reviewers may use AI to help: checking code for errors, reconciling outputs across runs, scanning a report for inconsistencies with the underlying results, or summarising long model documentation so the reviewer can focus.

Both are reasonable in principle. The concern is a loop in which AI-produced work is checked by AI, and the human reviewer's role shrinks to confirming that the tool raised no flags. Independent review exists to bring a second mind to the judgements in the work, which is the part an AI tool is least able to supply.

The duties that apply

APS X2 v1.1, "Review of Actuarial Work", in force from 30 January 2026: members must consider whether Work Review, and in particular Independent Peer Review, is appropriate and proportionate, weighing complexity, significance, likelihood of error and the degree of judgement (paras 1.1 to 1.4). See the APS X2 text.
The IFoA data science and AI guide asks members to consider independent peer review under APS X2 and to be cautious about using AI within reviews (2.35 to 2.38).
TAS 100 P5.2: model governance; the FRC Models guidance suggests additional governance for AI models (4.9).
Code Principle 2: competence and care, which applies to the reviewer as well as the author.

APS X2 does not mention AI, but each of its factors is affected by it: AI models add complexity, opacity raises the likelihood of undetected error, and model choices involve judgement.

What can go wrong

Illustrative scenario: a consultancy's reviewer is asked to check a reserving exercise that used a new machine learning method. Short of time, the reviewer uses an AI tool to compare the code against the method description and to check the report for internal consistency. The tool reports no issues. The review file records "reviewed, no exceptions". A year later the reserve proves deficient because of a data issue the tool was never asked about. When the client complains, the review file shows little evidence that anyone challenged the method or the data.

The problem is not that AI was used, but that the review did not address the judgements that mattered. See AI in reserving and LLM-written code and spreadsheets.

Why review matters for PI

In a negligence claim, the question is commonly whether the work met the standard of a reasonably competent actuary. Evidence that the firm identified a piece of work as high risk, arranged appropriate review and recorded the challenges raised is powerful. A review that is thin, or that delegated the thinking to an AI tool, is weak evidence and may itself be criticised. PI insurers may ask about review procedures when underwriting actuarial firms. If a review after the event finds a possible error in past work, check whether you need to notify a circumstance to your insurer. See PI for actuaries and what to document about AI use.

Practical controls

Treat material AI use as a factor pointing towards Independent Peer Review under APS X2.
Brief reviewers on where AI was used in the work and what it did.
Let AI support mechanical checks, but have the reviewer form their own view of methods, data and assumptions.
Record what the reviewer challenged and how it was resolved, not just a sign-off.
Do not put confidential client material into unapproved AI tools during review.
Review AI-drafted communications too; see AI-drafted actuarial reports.

How Apex helps

Apex arranges PI for actuarial consultancies, and a broker can help you present your review procedures to insurers at renewal. See the AI in actuarial work overview and the FRC and IFoA guidance explained.

Sources

Frequently asked

Does APS X2 mention AI?

No. APS X2 v1.1 does not mention AI. It requires members to consider whether Work Review, and in particular Independent Peer Review, is appropriate and proportionate, weighing complexity, significance, likelihood of error and judgement. AI use often raises those factors.

Can a reviewer use AI to check actuarial work?

Yes, for support such as code checks or consistency scans, but the IFoA guide advises caution about using AI within reviews. The reviewer should form their own view of the methods, data and judgements and record the challenges raised.

How does peer review help if a claim is made?

A documented, proportionate review is evidence that the firm took care, which helps show the work met a competent standard. A thin review, or one that relied on an AI tool without real challenge, offers much less support and may itself be criticised.

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.