FCA authorised · FRN 724952 0117 325 0027Quote & buy →
Aipedia: Actuaries
Pension actuaries can use AI to cleanse member data and support valuations, but they remain responsible for the data judgements and results under TAS 300 and APS P1. A small data error can flow into funding plans and member benefits for years.
Part of: AI in Actuaries
In short
Scheme data is often incomplete, with missing dates, duplicate records and inconsistent salary histories. AI tools can match, fill and flag records quickly, but every automated correction is a data judgement that TAS 100 and TAS 300 expect the actuary to understand. Member data is personal data, so confidentiality and consent rules apply, and the IFoA warns against entering data into third-party generative AI tools. Errors can lead to claims from trustees and, in some cases, sponsors. Actuarial PI usually responds subject to the policy terms, but duty of care to non-clients is a key question.
Before a valuation, scheme data has to be checked and cleansed. AI and rule-learning tools may be used to match members across administrator files, identify duplicates, infer missing data such as dates of birth or marital status, flag outliers in pensionable salary and reconcile movements since the last valuation. Language models may also be used to read scheme documents and summarise benefit rules. Further along, AI can help with running valuation scenarios and drafting results for trustees.
The attraction is time. The risk is that the tool makes thousands of small decisions that no one reviews individually, and some of those decisions determine liabilities. A defined benefit liability is only as good as the member data behind it, and errors in that data tend to persist from one valuation to the next unless someone goes looking for them.
Member data is personal data. If health data is used, for example on ill-health retirements, it is special category data. See actuaries, AI and client data confidentiality.
Illustrative scenario: an AI matching tool merges two deferred members with the same name and similar dates of birth, and fills a gap in another member's service history by inferring it from colleagues. The valuation proceeds. Later, when one of the members retires, the administrator finds the record wrong, and a review shows similar errors elsewhere. The trustees question the valuation result and the recovery plan agreed with the sponsor, and the costs of a data rectification exercise follow.
Other risks: an AI summary of the benefit rules that misses an amendment, inferred marital status feeding spouse's pension liabilities, and data entered into an unapproved chatbot. Mortality assumptions are a separate area; see AI mortality and longevity modelling.
A claim from trustees that negligent data work or valuation advice caused loss would usually fall within actuarial PI, subject to the policy terms. Sponsors and members are not normally the actuary's client, so whether a duty of care is owed to them depends on the facts and the engagement terms. Claims connected to pension work can surface long after the valuation, so continuity of cover and run-off matter. A data protection breach involving member data may also raise questions for cyber cover; see data breach liability for professional firms. For limits, see actuaries and PI requirements.
Apex arranges PI for pension actuaries and consultancies; a broker can check how your cover treats AI-assisted data work and whether cyber cover is needed alongside it. See the AI in actuarial work overview.
You can, but treat each inferred value as an assumption. Record what was inferred and how, test a sample against source records, and disclose the extent and effect in your report. TAS 100 and TAS 300 expect you to understand data limitations and communicate them to trustees.
Only if the tool is approved and covered by proper data protection terms. Member data is personal data, and the IFoA guide says care should be taken when using third-party generative AI tools because of privacy and other risks. Unapproved tools can create a data breach as well as a professional issue.
Possibly, but the sponsor is not usually the actuary's client, so a claim would turn on whether a duty of care arose on the facts and what the engagement terms say. Trustees, as client, are the more common claimant. PI usually responds to such claims subject to the policy terms.
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 callbackApex 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.
