Key Takeaways
- • APG, which manages €639 billion in AUM and serves as in-house manager for the €533 billion Stichting Pensioenfonds ABP, says AI adoption requires organizing workflows before automation and assigning a clear human owner, with René Steenhart stating "Without clear ownership, you get nothing."
- • APG is pursuing a "minimum viable governance" model in which governance scales with risk and application volume rather than adding bureaucracy, allowing staff to experiment without waiting for repeated approvals.
- • PGGM (€271 billion, in-house manager for Pensioenfonds Zorg en Welzijn) reports 100% of employees use Microsoft Copilot and 80% use its internally built PIA tool, with adoption contingent on employees trusting the technology and seeing its benefits.
- • Ford Motor Company's pension team built an internal AI chatbot for employee pension questions, and found that training it on real questions from past internal messages and emails improved its quality compared with relying on formal documents alone.
- • Tey Bannerman identifies three emerging AI risks for institutions: delegating key decisions to AI, optimizing for easily measured metrics over long-term value, and group-think from shared chatbot outputs.
Summary
At the Pensions & Investments World Pension Summit in The Hague on November 3, 2026, pension fund leaders and industry experts described how their organizations are deploying AI internally and for plan members, according to a report by Pensions & Investments. The central message was that the technology delivers value only when workflows are redesigned, governance is proportionate, and employees and members actually adopt the tools.
René Steenhart, managing director for digitalization and resilience at APG, said firms must organize workflows before automating them and must assign a human owner to every AI project. APG, which has €639 billion in AUM and manages the €533 billion Stichting Pensioenfonds ABP, is pursuing what Steenhart called a "minimum viable governance" approach. Under this model, governance grows with the risk and volume of applications and the number of agents being developed, giving staff room to experiment without awaiting repeated approvals. Steenhart stressed that this does not mean lower standards.
Tey Bannerman, a former McKinsey partner now working as an independent AI adviser, cautioned that new risks will emerge as AI becomes embedded in daily workflows. He cited the risk of people handing key decisions to AI, the tendency to optimize for what is easily measured rather than long-term value, and a group-think effect when many users receive similar answers from the same chatbots. He noted that many employees remain reluctant to delegate key decisions to AI.
Hens Steehouwer, chief innovation officer at Ortec Finance, which works with pension funds and institutional investors, highlighted AI's capacity to run advanced portfolio analysis quickly, combining scenario testing with strategic allocation and enabling new evaluation methods such as three-dimensional modeling of risk, return and factors including climate.
On adoption, Abe Scholten, innovation manager at PGGM, said employees must trust the technology and see its benefits before they will use it. PGGM pairs general-purpose tools for tasks such as email sorting with specialized AI tools for workflows, including within the investment process. Scholten said 100% of PGGM employees use Microsoft Copilot and 80% use PIA, an internally built tool similar to Copilot, and that building applications has become easier because teams can iterate quickly with AI.
For member-facing applications, Oliver Payne, reward director for continuous improvement and international pensions at Ford Motor Company, described an AI chatbot his team built to answer employee questions about pensions and retirement plans. It was based on existing HR documents and a new FAQ compiled from prior internal pension messages and emails. Payne said incorporating data on what members actually ask, and how they phrase it, improved the chatbot, and that without that material the tool gives "actuarial answers" that lack practical context.
René Steenhart, managing director for digitalization and resilience at APG, said firms must organize workflows before automating them and must assign a human owner to every AI project. APG, which has €639 billion in AUM and manages the €533 billion Stichting Pensioenfonds ABP, is pursuing what Steenhart called a "minimum viable governance" approach. Under this model, governance grows with the risk and volume of applications and the number of agents being developed, giving staff room to experiment without awaiting repeated approvals. Steenhart stressed that this does not mean lower standards.
Tey Bannerman, a former McKinsey partner now working as an independent AI adviser, cautioned that new risks will emerge as AI becomes embedded in daily workflows. He cited the risk of people handing key decisions to AI, the tendency to optimize for what is easily measured rather than long-term value, and a group-think effect when many users receive similar answers from the same chatbots. He noted that many employees remain reluctant to delegate key decisions to AI.
Hens Steehouwer, chief innovation officer at Ortec Finance, which works with pension funds and institutional investors, highlighted AI's capacity to run advanced portfolio analysis quickly, combining scenario testing with strategic allocation and enabling new evaluation methods such as three-dimensional modeling of risk, return and factors including climate.
On adoption, Abe Scholten, innovation manager at PGGM, said employees must trust the technology and see its benefits before they will use it. PGGM pairs general-purpose tools for tasks such as email sorting with specialized AI tools for workflows, including within the investment process. Scholten said 100% of PGGM employees use Microsoft Copilot and 80% use PIA, an internally built tool similar to Copilot, and that building applications has become easier because teams can iterate quickly with AI.
For member-facing applications, Oliver Payne, reward director for continuous improvement and international pensions at Ford Motor Company, described an AI chatbot his team built to answer employee questions about pensions and retirement plans. It was based on existing HR documents and a new FAQ compiled from prior internal pension messages and emails. Payne said incorporating data on what members actually ask, and how they phrase it, improved the chatbot, and that without that material the tool gives "actuarial answers" that lack practical context.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.