Asset managers and the use of AI: the AFM identifies opportunities and risks in its recent report

The report presents the findings of a survey among 323 Dutch asset managers on their use of AI, the associated risks and the current state of their AI governance. The report is relevant because it highlights not only the growth of AI adoption, but also the AFM's observation that governance and investments in AI-applications frequently lag behind.

AI adoption grows, but unevenly across the sector

The report reveals that over half (53%) of Dutch asset managers are either using AI or plan to do so within the next year.  It should be noted that the AFM uses "asset managers" as an umbrella term throughout the report, which includes not only investment firms, fund managers and depositaries, but also proprietary traders and trading venues.

Across the various licence categories, proprietary traders (72%) and Organised Trading Facilities (OTFs, 75%) lead in adoption. In general, the AFM observed that larger asset managers (measured by Assets under Management) consistently use AI more than smaller firms.

Despite this growth in adoption, investments in AI generally lag behind the pace of deployment. Some 71% of asset managers indicate that they do not have a specific budget for AI use. The AFM explains this in part by the fact that many firms rely on publicly accessible AI applications (such as ChatGPT or Microsoft Copilot) or on general budget allocations. In contrast, a small minority of asset managers invests more than EUR 1 million per year. Overall, the AFM notes that adoption is growing and that a majority (60%) of respondents expect to increase AI spending over the coming two years.

Added value for asset management practice

Among the respondents, AI is currently mainly deployed for information gathering and supporting decision-making and analyses. Beyond this, AI is also primarily used for writing research reports and analysing unstructured datasets. Proprietary traders in particular stand out in their application of AI solutions: a large group uses AI to optimise trading algorithm parameters (65%), improve trading strategies (62%) and predict price movements or market trends (62%). Proprietary traders also stand out by their preference for Machine Learning techniques such as ‘supervised’ and ‘unsupervised’ learning.

For these practices, nearly all asset managers (94%) rely on General Purpose models like ChatGPT, Claude and Google Gemini. About 18% of respondents use custom models developed internally, and 14% rely on [other?] external providers.

Governance and risks associated with AI

The AFM emphasises that the use of AI by asset managers can be of significant practical value. The AFM observes that AI can contribute to improved portfolio allocations and market analyses. Nevertheless, the AFM also identifies considerable risks and notes that a proportion of asset managers are insufficiently prepared for these. For example, the majority of asset managers make use of commercial cloud solutions, and the AFM calls for particular attention to the handling of data and data privacy, as well as the operational dependency that may accompany this.

One in four respondents lacks any policy (and has not started developing one) governing or limiting employee use of AI. As for Generative AI specifically, only 28% have adopted a formal policy. Furthermore, the majority of those surveyed have not implemented a dedicated ethical handbook or code of conduct addressing the use of AI within the organisation. On the specific use of automated AI-agents, the AFM highlights that just 2% of respondents have developed a formal policy.

These differences in policy adoption are also evident with regard to the training of employees in the use of AI. Fewer than half of asset managers provide general awareness training for all employees on AI (including the use and functioning of AI systems). Only fifteen asset managers indicate that they have established specific advanced training programmes for AI developers, of which four are proprietary traders. In the context of ‘AI literacy’, the AFM points out that organisations that develop or use AI have been required since February 2025, under the EU AI Act, to ensure that their staff possess sufficient knowledge and competence to deploy AI in a responsible manner.

Outlook

The AFM's findings are similar to what DNB reported earlier this year in respect of the use of AI by insurers (see link to our previous publication). Both the AFM and DNB note that larger firms are further ahead in AI adoption, while smaller market participants struggle to keep up. Read together, the two reports send a clear message to the Dutch financial sector: adopting AI without investing equally in sound policies and monitoring will be scrunitised. 

The AFM has confirmed that it will continue to observe how asset managers deploy and control their AI applications in the period ahead.