Brighter Consultancy Blog

AI in Actuarial: The Opportunity Isn’t Just Faster Modelling

Written by Sarah Watkins | Nov 20, 2026, 1:05:43 PM

When discussed in an actuarial context, much of the attention around AI naturally falls on modelling. Whether models can run faster, process larger datasets or automate activity becomes the topic of conversation.

There is certainly potential for AI to do all of those things, but some of the more immediate opportunities sit across the wider workflow. Actuarial teams spend significant time preparing data, validating outputs and reviewing changes, and AI is currently supporting several of these activities. With this, firms now have an opportunity to look at how the entire process works rather than focusing solely on model runtime.

Recent research suggests the shift is already underway. Earlier this year, the Society of Actuaries’ AI Survey found that around 60% of respondents across both experience groups were using enterprise AI models provided or licensed by their organisations. The research also found growing interest in incorporating AI into day-to-day actuarial workflows, evaluating and validating its output and understanding how it can be used responsibly.

As actuarial teams increasingly look to AI to improve their workflows, the question becomes one focused on how AI can remove friction from actuarial work while maintaining the controls and professional judgement the function depends on.

The opportunity goes further across the modelling process

Before a model is run, information needs to be gathered, cleaned and prepared. Once results are produced, actuaries need to test them, understand movements, document changes and determine whether the output is viable.

Previous research from the Society of Actuaries has documented the practical uses of generative AI across the workflow, including pre and post-model data processing, generating test cases, testing documentation and calculations, and supporting model validation.

Often absorbing a lot of capacity across teams, these actions still require manual processes and repeated checks in many organisations. Using AI to support them could shorten the overall process without changing the responsibility actuaries have for the final result, reducing some of the work required to reach a point where specialist judgement can be applied.

Validation could become more efficient

Validation is a key area for actuarial teams to explore, as it already involves reviewing large volumes of information and identifying where something has changed or does not behave as expected.

AI can be particularly useful in supporting the generation of test cases and comparing outputs, especially when teams are reviewing model changes, migrations or updates to assumptions.

However, it also creates a different set of responsibilities; a result produced or reviewed with AI still needs to be understood by the actuary responsible for it. Teams also need to know what information the tool has used, whether the output can be reproduced and where additional review is required.

The Society of Actuaries has already published practical guidance on governance and ethics for generative AI models, covering areas including oversight, validation and the risks created by less transparent models. As such, insurers introducing AI into validation should take the same disciplined approach applied to other actuarial tools, with controls reflecting the significance of the output.

Documentation and reporting can also change

Actuarial teams produce a considerable amount of supporting materials around their work, such as assumption papers, model documentation, management commentary and explanations of movements, all of which require specialist input.

Used appropriately, AI can help teams produce an initial draft, organise technical information or compare documentation between reporting periods. However, the actuary still needs to review results carefully, particularly where documents explain assumptions, model limitations or material financial movements.

The use of AI in drafting also has implications for how actuarial information is communicated outside the function, as technical results often need to be understood by finance teams, underwriters, risk functions and senior leaders who do not work directly with the model.

AI can help to translate detailed analysis for different teams, but the underlying message still needs to be controlled by an actuary. To this end, IFoA’s earlier research into generative AI adoption highlighted both report generation and stakeholder communication as areas where actuarial workflows were already beginning to change.

Better use of information could improve actuarial challenge

The challenges faced by actuarial teams are rarely limited to producing another calculation; why experience has moved, comparing assumptions, investigating an unexpected result or identifying a pattern across several sources all require time.

AI can make analysing large volumes of information easier, which could be particularly useful in areas such as experience analysis, reserving or pricing, where teams regularly need to understand movements across multiple periods or identify which information requires closer attention.

However, the value here depends heavily on the quality of the underlying data; AI cannot compensate for poorly defined information or inconsistent inputs, and actuarial teams still need enough understanding of the source data to recognise when an apparent pattern is misleading.

If a firm is already investing in stronger actuarial data and systems, AI could provide another way to make better use of those foundations.

Professional judgement remains at the forefront

Rather than removing the need for actuarial judgement, further use of AI makes the point at which it is required much clearer. The latest SOA member research found that professional judgement remained central to respondents’ views of AI even as engagement increased and governance became more structured.

An AI tool can identify an unusual movement or produce an initial analysis of results, but actuaries still need to consider whether that explanation is credible within the context of the portfolio, model and wider business.

They also need to recognise when further investigation is needed, which is closely aligned with the direction the IFoA has taken in its 2026 AI Manifesto; IFoA argues for an “actuary-in-the-loop”, drawing on actuarial expertise in modelling, risk, governance and uncertainty to strengthen how organisations develop and use AI.

The wider workflow should still determine where AI is used

When considering how AI can support activity, insurers should look beyond the question of which modelling tasks can be automated. As a start, it is useful to map where time is currently being spent across the entire process; data preparation, validation, documentation, analysis and communication can all create bottlenecks, particularly where activity remains manual or relies heavily on a small number of specialists.

The strongest use cases for AI are likely to be those where the problem is already understood and teams can clearly explain what improvement they expect to see. In practice, this could mean reducing the time required to prepare model inputs, improving the efficiency of validation or making it easier to produce documentation. In another function, the benefit may come from helping actuaries analyse more information before making a recommendation.

Governance also needs to develop alongside adoption, particularly as the SOA’s latest research shows that regulatory and compliance risk, skills gaps and unclear return on investment are barriers to developing the use of AI.

The opportunity for actuarial functions is much broader than making models run faster. AI also has the potential to change the work that happens before and after the model, reducing some of the friction around technical processes and giving actuarial teams more capacity to understand, challenge and communicate the results they produce.

Brighter Consultancy works with insurers and reinsurers on actuarial transformation, including modelling, model harmonisation and re-tooling, technical solutions, platform selection and systems migration.

If you are considering how AI could support your actuarial function, speak to our team about where it could have the greatest practical impact.