Risk management and AI

Ms Begass and Mr Kleijs (Accenture) explained what the rise of artificial intelligence (AI) means for pension funds, administering organisations and asset managers.

Accenture is a global technology consulting firm with approximately 780,000 employees worldwide, 8,000 of whom are engaged in risk management and compliance.

Three primary forms of AI are distinguished:

• Classical AI, focused on pattern recognition, works with structured data and requires a lot of assumptions. This type is used to support the ALM function;
• Generative AI uses unstructured data, facilitates the creation of policy documents and interprets complex issues. Think of applications such as chatbots or support within risk management processes;
• Agentic AI can support or even take over processes, makes connections between various data and systems, and controls other AI agents within complex processes. This type of AI is suitable for decision-making and compliance, if adequately trained and monitored.

The use of AI increases the efficiency and speed of processes, resulting in a structural improvement in productivity. For example, AI monitors data quality more effectively than humans and helps reduce execution costs, allowing employees to spend more time on substantive tasks.

Within risk management, AI determines how much risk the asset is exposed to, supports allocation, management, administration and monitoring.

Directors and boards should carefully consider how AI can be used in pension schemes, asset management and policy preparation. A strategic approach around AI is necessary, including integration into governance and review of staff deployment.

Fewer staff may be needed, but there is still a need for critical monitoring of the use and results of AI. The rise of AI is also leading to new features. Professionals with knowledge of investment processes can interpret AI results, while it becomes more difficult for young people to find entry-level positions; these tasks are increasingly performed by AI. This reduces their chance of learning business processes and asset management and assessing AI outcomes. Mentorship is essential, as experienced experts take on a new role as assessors of AI outcomes.

Compliance with laws and regulations governing the use of AI

Mr. Van den Bergen went through the laws and regulations that apply to the use of artificial intelligence by financial institutions.

Artificial Intelligence is increasingly being used in the financial sector, including pension funds and asset managers. AI can support decision-making, but it also poses new risks. Think of bias in models, incorrect output, lack of explainability or reputational damage in the event of incorrect use.

Regulators such as Dutch Central Bank and the Financial Markets Authority therefore expect organisations to actively control the use of AI and to be able to explain how AI systems work, why they are used and how the risks are mitigated.

The European AI Act introduces a new legal framework for this. AI systems are classified into different risk categories, ranging from prohibited applications to high-risk systems. High-risk AI is subject to obligations regarding risk management, data quality, documentation and monitoring, among other things.

Anyone who develops an AI system and uses it under their own name can be regarded as a provider under the AI Act. This entails heavier compliance obligations than when an organisation is marked as a user. For pension funds and asset managers, this means that they must already take steps to use AI responsibly.

Practical first steps include:

• An organisation-wide inventory of all AI applications;
• Classification of these applications according to the AI Act;
• Determine whether the organisation is a user or provider;
• Setting up governance, supervision and monitoring;
• Involving the board in AI strategy and risks.

The pension fund also remains ultimately responsible for outsourcing. Therefore, insight is needed into the use of AI by asset managers and other service providers.

Contracts should be amended where necessary with agreements on transparency, incident reports and audit rights. The most important message for directors is that AI is not just a technological development, but a new governance and supervisory issue. Boards must understand how AI is used within their organisation and actively manage the associated risks.

Pension communication and AI

Mr. Aboikoni and Ms. Van Dam discussed the strategic choice for the use of artificial intelligence in pension communication, and implementation.

AI can be used as a means for effective customer (participant) service (CX). This is useful, because nowadays we often see that pension organisations have to deal with a resource limitation in the field of CX. This, while the expectations of participants regarding customer service are higher than ever.

Strategy

KPMG uses the model of The Six Pillars. Together, they make up the customer experience. The Six Pillars are:

• Personalization;
• Time and effort;
• Expectations;
• Resolution;
• Empathy;
• Integrity.

In this model, the pension fund chooses one Pillar in which the pension fund wants to be the best, one Pillar is chosen in which the pension fund distinguishes itself from all others, and the remaining Pillars are equal to what other pension funds do. Subsequently, the chosen pillars that the pension fund wants to focus on can be used at as many touchpoints in the customer journey as possible. AI can play a role in this. It is important that cooperation is sought at an early stage with, among others, a process expert, AI expert, participant representatives and a compliance/lawyer, to discuss the following points:

• What part of the participant's customer experience will go without any human cooperation;
• Which risks are identified; 
• What value is reached.

Application of AI to service plan participants

An application of AI is the digital assistant. This digital assistant knows everything about pensions, recognises the plan participant and his or her degree of financial literacy. AI can be used to segment the participant base. The portal then looks different for the different types of plan participants. No more standard portal for everyone.

Another application is that AI can predict behaviour. For example, whether someone is more at risk of a pension gap, or will not respond to communication. A pension fund can then act proactively instead of repairing afterwards.

Conclusion

Thanks to AI, more quality is possible in plan participant service. The quality of the operation improves and there may be cost savings in the implementation. 

AI Beyond the Hype

Mr. Cartwright discussed artificial intelligence as a sector to invest in.

Introduction

AI investing has shifted from early enthusiasm to a deeper look at economics, capital intensity, and returns. Demand for AI remains strong, and capital investment by AI companies is accelerating—investment in data centres alone could reach $4.5–5 trillion over the next five years. However, power availability, access to the power grid and construction timelines are becoming critical limiting factors. AI has entered an industrial phase. Investment success now depends less on exposure to the theme of AI and more on execution — specifically, which AI companies can deploy capital efficiently, secure scarce inputs, sustainably finance growth, and turn investments into sustainable profits.

Investment aspects

Despite high spending by AI companies, the returns are not yet uniform. Productivity gains occur gradually, and outcomes increasingly depend on data quality, operational capacity, balance sheet strength, and pricing power. As with previous waves of technology, high investment does not guarantee high profitability. The winners are those who can generate returns above their cost of capital.

AI is also expanding across industries, reaching beyond software and semiconductors to energy, utilities, data centres, capital goods, materials, and real-world assets. This creates a diversification problem: portfolios can appear to be sector diversified while still being highly exposed to the same underlying AI infrastructure.

Investing in AI for pension fund portfolios

You can't just look at the sector level. Look at the spread within the tech sector. Software and IT services are feeling the effects of changing expectations and adaptations to growth profiles. The investment implication is that selectivity matters. The next phase of AI investing will be to focus on companies that manage constraints, maintain financial flexibility, maintain earnings stability, and demonstrate clear returns on capital. For portfolios, this argues for measured, diversified exposure across the AI value chain, with an emphasis on valuation discipline, resilience and long-term cash flow generation, rather than broad thematic exposure.

Conclusion

• AI remains a powerful long-term structural force;

• The debate has shifted from possibility to profitability;

• Execution, capital discipline and economic exposure determine who ultimately reaps the revenues.

Global stock selection with AI

Mr. Susinno discussed  how Pictet Asset Management uses AI in an innovative way as a means to achieve additional returns in equity investments.

To this end, Pictet has developed an AI-driven quantitative equity strategy that takes advantage of short-term market inefficiencies. This global enhanced indexing equity strategy has a low tracking error and a low fee.

The strategy has the characteristics of an equity index, with very small country and sector deviations, similar factor exposures and with a beta of 1 compared to the index. With about 400 data characteristics, an AI model is trained to predict returns for a 1-month period. These are quantitative data such as sentiment, market activity and fundamental factors. Based on the different predictions in returns, the shares get a small overweight or underweight in the portfolio compared to the index.

Pictet has launched the AI-driven equity fund almost 2 years ago. The track record shows that value is added by linear combinations of factors, but also by interactions between the characteristics, which cannot be exploited by linear models.

Pictet will release regional versions (Europe and U.S.) and an emerging markets AI-driven equity strategy in the course of 2026.