AI is emerging as a strategic opportunity for Swiss independent asset managers, but its adoption is still largely at the experimental stage. A study conducted by the Lucerne University of Applied Sciences and Arts amongst asset management firms highlights the main barriers to its roll-out, ranging from data quality to privacy protection. Tatiana Agnesens and Manfred Stüttgen emphasise here the crucial role of a clear strategy in moving from initial applications to the full integration of AI into asset management practices.
Half of the asset managers you surveyed are already using AI, but only 11 per cent of use cases have progressed to the stage of large-scale implementation. What is preventing all these managers from moving beyond the experimental stage?
EAMs are very interested in the potential use cases for AI. However, they remain concerned about data protection and privacy when it comes to integrating AI into their internal processes.
They also remain fairly sceptical about the reliability and accuracy of the results produced by AI. The main obstacle to large-scale implementation, however, remains the lack of a more comprehensive AI strategy. When an EAM decides to make AI a priority and to take full advantage of it, these various obstacles can be overcome.
Only 27 per cent of EAMs consider their data to be sufficiently structured for AI. By contrast, 83 per cent of them see AI as a strategic opportunity. To what extent is data preparation likely to become a hindrance?
Indeed, data preparation is currently regarded as the main barrier to AI adoption by EAMs. They believe that their data is neither sufficiently structured nor reliable enough to produce the expected results. Addressing these issues therefore forms the foundation for the successful adoption of AI. Some EAMs have already begun this work, others are in the process of doing so, and we expect more of them to follow suit in the near future.
Investment research is currently one of the areas where AI is making the most progress, whilst onboarding and data reconciliation lag behind despite having comparable potential. How do you explain this disparity between processes with a high text-based component and those more closely linked to workflows?
Investment research is one of those text-heavy processes in which AI can utilise large volumes of external data. Generic and familiar tools such as ChatGPT can also be used. Share screening, as well as qualitative and quantitative analyses, can be carried out very effectively without raising any particular issues relating to data or confidentiality.
Onboarding and data reconciliation follow a different logic. They require the implementation of specific tools, based on internal projects and a dedicated budget, ideally as part of an enterprise-wide AI strategy. However, such a strategy is still often lacking.
Data protection concerns were cited by 85 per cent of respondents, whilst regulatory uncertainty decreases significantly with experience. Does this mean that FINMA’s AI recommendations are being adopted more quickly than anticipated?
To our surprise, we found that FINMA’s AI recommendations are not regarded as a major obstacle to its implementation. EAMs must, of course, take them into account, but they see themselves as experienced and risk-averse players, capable of complying with such recommendations.
They view the implementation of AI in the same way as any new technology, with a number of requirements to be met, including appropriate governance, controls to ensure data quality and adequate documentation. Data protection is an important element of this framework, but it stems less from FINMA’s requirements than from EAMs’ desire to safeguard their clients’ confidentiality.
You say that EAMs are looking for “a strategy and a roadmap, rather than more tools”. What might a realistic roadmap look like for a medium-sized EAM today?
It could, in particular, take the form of a practical guide to AI for EAMs, accompanied by an overview of potential use cases, how to implement them and the associated costs. It would also be useful to include a quantitative and qualitative assessment of the risks and benefits. This would give EAMs a more concrete picture of the costs, benefits and opportunities offered by AI.
Looking at the value chain as a whole, which processes currently show the greatest gap between their AI potential and their level of adoption?
Data quality and reconciliation, as well as onboarding and account opening, are among the processes where AI potential is high but adoption remains relatively low. This is where the gap is currently most pronounced.
Conversely, many EAMs have made progress in investment research, proposal generation and client communication. They have already narrowed the gap between AI’s potential and its adoption.
Between recruiting in-house AI expertise and using external consultants, which approach should Swiss EAMs prioritise over the next two years?
We believe that more and more EAMs will turn to external consultants. This is a simpler way to quickly acquire the necessary skills. However, to integrate AI into the organisation on a long-term basis and strengthen its internal capabilities, it would also be advisable to develop skills in-house. More in-depth training in AI could contribute to this objective. Indeed, many EAMs have expressed this need.
Detailed results of the study
AI Use Cases in External Asset Management (EAM) — Finup Research