AI in asset management

For its report AI adoption and trends in securities markets, ESMA asked firms across the EU about their AI use: 395 firms reported more than 800 use cases. Drafting and summarizing text was the largest group, followed by internal assistants and code generation. Portfolio risk management, portfolio optimization and algorithmic trading with AI were rare, and most of those cases were still in development or testing. 87 percent of all use cases served internal work, 10 percent client relations and 3 percent the investment services themselves.

In practice an asset management firm uses AI in research first: a model reads annual reports, earnings call transcripts and broker notes and drafts a summary for the analyst. Next come the operations: fund accounting, reporting and the back office, where Helaba Invest, for example, processes incoming invoices for its funds with AI. The portfolio manager still makes the investment decision, and the compliance team documents who approved what.

A fund can use AI in its process, or it can invest in AI companies. ESMA looked at both in its article Artificial intelligence in EU investment funds. Funds that promote AI in their investment process remain a small share of the industry. They have not delivered significantly higher or lower performance than others and have seen outflows. Asset managers use generative AI and large language models mainly to support decisions made by people. At the same time, EU funds hold more and more shares of AI-related companies.

A retail investor who searches for an AI fund usually finds the second kind. The Deka-Künstliche Intelligenz fund (in German), for example, is an actively managed equity fund that invests in AI infrastructure, AI applications and companies that profit from AI. Whether a fund manager uses AI to run the portfolio is a separate question, answered in the fund documents.

Union Investment, the fund company of the cooperative banks, started one of the first AI-based fund portfolio services in Germany together with LAIC Vermögensverwaltung, a subsidiary of LAIQON. According to LAIQON (in German), the service, called WertAnlage, combines human expertise with AI in the investment process: clients set preferences on risk, asset classes, regions, sustainability and themes, which allows more than 2,600 individual allocations, and the AI proposes the exact allocation after analyzing more than 125 million data points every day. The service is aimed at wealthy private clients.

A service like this is portfolio management under MiFID II, so the suitability test and the cost disclosure apply as they would with a human manager. The robo-advisor page covers the license side, and the asset management in Germany page covers the KAGB authorization every German fund manager needs.

The EU AI Act does not list investment advice or portfolio management as high-risk. The duties that matter come from financial law: ESMA's public statement on AI in investment services says that MiFID II applies in full, that a firm must act in the client's best interest and be open about the role AI plays, and that it must manage risks such as algorithmic bias, poor data quality and a lack of transparency. ESMA also warns that text generators can "hallucinate", which in advice or portfolio reports can mislead a client.

Asset managers that run models on a cloud platform add a third layer: the Digital Operational Resilience Act, which requires a register of every ICT service contract and exit plans for providers behind critical functions. A market data vendor with an AI layer on top is such a provider. The EU AI Act in financial services page explains where the high-risk line runs.

Finance Loop connects portfolio managers, quants, data engineers and compliance people at asset managers who put AI to work. Finance Loop connects the finance, IT and AI communities in Germany, Austria and Switzerland, with events in Frankfurt, Munich, Berlin and Hamburg.

Finance Loop was a partner of AI in Finance Paris, organized by Finteda for people from wealth management, market data and quantitative research, with speakers from BNP Paribas, Deutsche Bank, Berenberg and RAM Active Investments. Finance Loop supported the AI & Financial Market Data meetup in Paris, where quantitative developers from Deutsche Bank met people from Amundi, FactSet and QuestDB. Finance Loop member Fayssal El Mofatiche co-wrote Building AI Agents for Finance, which builds agents for fundamental analysis, deep research and trading.

Asset managers mostly use AI for internal work. In the ESMA survey the largest groups were drafting and summarizing, internal assistants and code generation. AI that optimizes portfolios or trades on its own was rare.

ESMA found that EU funds that promote AI in their investment process have not delivered significantly higher or lower performance than other funds, and they have seen outflows. They remain a small niche of the market.

The AI Act does not name portfolio management or investment advice in its high-risk list. MiFID II applies in full, though, and a client chatbot must tell people they are talking to an AI.

The CFA Institute Research Foundation published AI in Asset Management: Tools, Applications, and Frontiers, which teaches practitioners to balance automation with human oversight. The hub answer on AI in wealth management summarizes the ESMA data next to it.