Artificial intelligence is changing pension provision faster than many expected. Between efficiency gains, personalized solutions and unanswered questions about liability and trust, a clear picture is emerging: the future is hybrid and challenging.
The momentum is unmistakable. What was considered an innovation just a few months ago already seems outdated today. Generative AI has developed from an experiment to an everyday tool in a very short space of time, including in the pensions sector.
One particularly impressive milestone shows that systems now understand the Turing test better than humans. AI not only communicates efficiently, but increasingly credibly. This shifts the discussion from “whether” to “how” and, above all, how quickly.
Seven fields of action and a clear focus
The discussion in this year’s first virtual Future.Talk of the Institute of Insurance Economics at the University of St. Gallen (HSG ) on the topic of “The impact of artificial intelligence on pension provision” made it clear where the industry currently stands. Seven central fields of action structure the development:
The highest level of approval was given to increasing efficiency. Automation in administration, processing and customer interaction is seen as a direct lever, albeit with limitations. Productivity increases, but real efficiency gains are not guaranteed.
The role of AI in financial education was rated almost as highly. Access to pension knowledge could improve fundamentally, provided that content remains understandable, neutral and trustworthy.
The trend towards a hybrid model is also clear. AI will not replace advice, but complement it. Routine will be automated, but complexity will remain human.
AI-supported investment decisions and fraud prevention were assessed more cautiously. Uncertainties dominate here, particularly with regard to model risks and regulatory issues.
Rethinking consulting: personalized and permanent
Dynamic pension provision is a key vision for the future. Instead of static products, adaptive systems will emerge that continually adjust to life situations. In future, career changes, starting a family or changing residence could be automatically recognized and translated into pension recommendations. Scenarios will be calculated in real time, savings plans adjusted and risks reassessed.
The vision: provision is no longer planned, but accompanied.
But this development brings new requirements. Data quality is becoming a decisive factor. Without clean, structured data, any AI will be ineffective; the so-called “garbage in, garbage out” principle applies more than ever.
Efficiency is not a sure-fire success
Many processes, such as document processing, claims handling and internal communication, can be accelerated. However, efficiency is not automatically achieved through technology. Simply digitizing processes without questioning them often only increases complexity.
The crucial point is that automation must take place with clear objectives, not for its own sake.
People remain central
Despite all the progress made, pension provision remains a business of trust. Especially in sensitive situations, such as disability, death or retirement, human interaction is indispensable. AI can analyze, structure and prepare, but the final classification, empathy and responsibility still lie with humans.
This is also confirmed in practice: customers are open to AI as long as they retain control. Fully automated decisions are accepted, but only within defined limits.
Data as a competitive advantage
A key learning from practice shows that the actual competitive advantage does not come from AI itself, but from the combination of technology and proprietary data. They all use similar models. The difference lies in the database, process understanding and internal expertise.
Companies that structure their data landscape and make it accessible create the basis for scalable AI applications. Without this basis, the potential remains limited.
Regulation lags and slows down at the same time
Regulation remains an area of tension. While Europe already has a framework in place with the AI Act, Switzerland is still more cautious. At the same time, it is clear that too much regulation can slow down innovation. Too little creates uncertainty.
In particular, the question of liability remains open. Who is responsible if AI makes wrong decisions? Currently, the provider is liable, but many detailed questions remain unresolved.
Agentic AI: The next step
A new phase is beginning with the development towards “agentic AI”. Systems no longer just act reactively, but autonomously. They plan, decide and execute processes, such as the automatic adjustment of pension plans or investments.
This development opens up enormous potential for efficiency, but also increases risks. Control, governance and a clear allocation of roles are becoming crucial.
Transformation with a sense of proportion
Pension provision is facing a fundamental change. AI will make processes more efficient, advice more personalized and knowledge more accessible. But the transformation is not a sure-fire success. It requires a clear strategic focus,
clean data structures, internal expertise instead of external dependency
and a consistent balance between technology and trust
The direction is clear: man and machine are growing together. The decisive factor will be how well this collaboration is designed.
The next Future.talk 2/2026 will take place on September 8 between 2 and 5 pm.
Binci Heeb
Read also: Pension provision between growth, security and AI
Search:
Sponsors:
