Dick Kamp & Raymond van Es: Applications of AI in the pension sector

November 18, 2025

Artificial Intelligence Pension system Pensionfunds Pension

_{This column was originally written in Dutch. This is an English translation.}

By Dick Kamp, Director of Pensions, Investment and Risk, Milliman Pensioenen, and Raymond van Es, Principal and Practice Leader Data Science & AI, Milliman

AI has the potential to radically change the pension sector. Pension funds, pension administrators and pension insurers face complex challenges, such as managing large amounts of data, accurately predicting future liabilities and returns, and optimising customer communication. AI offers increasingly innovative solutions for these challenges.

There are already many opportunities to apply AI in the pension sector:

Risk prediction and asset management: AI models can analyse enormous amounts of financial and economic data to predict the risks and returns of different investment portfolios. This enables pension funds to optimise their investments and respond better to market risks.
Personal communication and customer support: Chatbots and virtual assistants help to respond to participants more quickly and personally. AI can also (if trained on information from specific schemes) generate personalised pension statements, answer questions and proactively inform participants about important choices. AI can also help to determine (provisional) individual risk profiles of participants by analysing open data.
Process automation: AI can automate administrative processes, such as processing changes, assessing pension applications and performing compliance checks. This reduces costs and also minimises the risk of human error.
Policy formulation: AI can support policy formulation. By asking specific questions to Large Language Models, AI can check for coherence and consistency on the one hand, and on the other hand, it can contribute additional elements that further strengthen the policy being developed.
Risk analysis: AI models are already particularly powerful in supporting front-line risk management. They can identify and analyse risks within a described case and propose appropriate control measures. This supports pension funds in conducting their business with integrity and control.

Challenges and control measures

Although AI offers many opportunities, there are also significant challenges:

Data quality and privacy: Pension funds work with sensitive personal data that can be many decades old. Ensuring data quality and privacy (GDPR) is essential. The quality, completeness and representativeness of the data used to train an AI algorithm determines the quality of the application.
Transparency and explainability: Decisions made by AI systems must be explainable to (former) participants, regulators and internal stakeholders.
Ethics and bias: AI models can unintentionally produce discriminatory outcomes if the underlying data is biased. This requires careful monitoring and ethical considerations. It is therefore important for pension funds to establish an ethical framework for the use of AI.
Dependency: Implementers, risk managers and directors may rely heavily on the outcomes of AI models and thereby unintentionally develop lower-quality policy-making. Carefully and analytically following the BOB process in a disciplined manner limits this risk and actually strengthens the administrative processes. Particularly in the decision-making phase, the AI output is systematically compared with additional data, professional expertise and ethical considerations, so that the final decision is based on a broadly supported, critically tested foundation rather than on (excessively) one-sided reliance on the algorithm.

Impact of the new pension law (Future Pensions Act)

The Future Pensions Act, which was passed in 2023, introduces a new pension system with more individualised accrual and greater transparency for participants. This requires more customisation, frequent communication and more flexible management of pension entitlements. This is a system in which AI comes into its own even more, because it is able to use large amounts of data to make predictions that help with ...

... personalising communication and scenario analyses for participants.
quickly processing changes in personal situations.
... modelling different future scenarios, so that participants can make better-informed choices....
predicting risk profiles in the context of “know your customer”.
... developing more differentiated investment profiles, enabling participants to make more personal investment choices and thus become more self-empowered while maintaining a constant personal (age cohort) risk profile.

And let's realise that we are only scratching the surface of the possibilities that AI will offer us.

The AI Pension of the future

In addition to optimising and supporting activities under the current pension system, AI could also lead to a new, innovative pension scheme in the long term. This AI-based pension scheme could function as a dynamic and personalised system, with AI agents proactively supporting participants in making choices about their pension accrual. These AI agents could perform complex analyses based on an individual's personal situation, market conditions and life goals, and thus provide personalised advice or automatically adjust the investment strategy. In addition, AI agents could support the collective – for example, a pension fund – by identifying trends and risks and optimising collective decisions, so that both the individual and the group benefit from smart technology.

However, it is important to realise that not all activities should be performed by AI. Certain processes, such as calculating statutory benefits or processing fixed administrative data, are deterministic in nature and can be handled more efficiently and transparently with traditional systems. Moreover, in some areas, the “human in the loop” remains essential, for example when making ethical decisions or guiding participants in making far-reaching choices. By using AI primarily where uncertainty and personal preferences play a role and keeping people involved in sensitive or complex situations, a pension scheme is created that combines innovation and relatively greater certainty. This leads to greater insight and involvement on the part of participants, while the collective as a whole benefits from smart support.

The importance of model validation

As AI models are increasingly used for critical decisions within the pension sector and the regulation of these models is also increasing, regular and careful model validation is crucial. This means that models are checked for compliance, accuracy, reliability and fairness. Validation prevents wrong decisions, limits risks and ensures trust among participants, directors and regulators. To validate AI models, a validation framework has been developed, in line with the steps for developing AI applications, which tests these models for compliance with both current regulations and functional specifications, i.e. does the model do what it was developed for? The quality of the management environment (design and existence) surrounding the model to be validated is also subject to validation.

Conclusion: the pension sector at a tipping point

AI is transforming the pension sector into a proactive, personalised service that can effectively respond to the challenges of the new pension law. Pension funds and pension administrators that are now investing in AI knowledge, developing an ethical framework and starting with well-validated applications are building a crucial lead. The future does not lie in replacing human expertise, but in a symbiotic system: AI takes over routine tasks and complex analyses, while professionals focus on trust-based relationships and strategic decision-making. This will create a pension sector that is technologically advanced, but where human input remains essential. Are you ready for this transformation?

This is the forty-fourth column in a series on risk management. The series aims to encourage readers to consider risk management as an integral part of running a pension fund