A New Era for Finance

Motivated by the rapidly developing capabilities of AI systems, Ralph S. J. Koijen, the AQR Capital Management Distinguished Service Professor of Finance and Applied AI and a Fama Faculty Fellow, organized a conference in February that brought together AI researchers, academics, and experts from the asset management industry to understand how these systems will reshape the ways markets are analyzed and risk is managed. The focus of the gathering was to discuss new applications, open questions, and the boundaries of what AI can achieve in finance.

Koijen convened a group of roughly 35 high-level experts from AI labs, AI-for-investing startups, the world’s largest asset management firms, and top academic institutions for the first of what he hopes will be an annual event.

“Everyone from fundamental managers to quant-market-maker types is deep, deep into AI now and exploring its possibilities and limitations,” Koijen said in a conversation afterward. “It’s not just experiments anymore. A lot of people are already building systems around them.”

Participants agreed that AI development tools, especially coding assistants, are advancing quickly, making it easier to build and explore complex AI workflows. Instead of writing every piece of code manually, researchers and others can now rely on AI-assisted development environments that help them construct data pipelines, orchestrate agents, and experiment with new architectures.

Koijen said the discussion focused in part on how much financial work can be automated as models improve. New tools are making it easier to build multiagent AI systems. Large language models can already summarize transcripts, extract themes, and analyze sentiment. The next generation of systems is going further.

Rather than using a single model to analyze a document, agentic systems can deploy specialized agents: one reading earnings calls, another parsing financial statements, a third tracking macroeconomic news. Those agents can then coordinate their outputs, producing a richer picture of companies and markets. As agentic workflows grow in complexity, the optimization, evaluation, and benchmarking of AI systems become increasingly important. This topic was discussed at length during the conference.

Researchers, technologists, and investors are beginning to converge on a shared realization: The most important breakthroughs do not come from better models alone but from how those models are integrated into research workflows. As those systems mature, they could begin to automate larger portions of the investment process itself.

For asset managers, the implications are significant. Research workflows that once relied on analysts to read and synthesize information are now more likely to involve AI systems performing the first layer of analysis. Similarly, modern AI systems can help in the design of quantitative models, including risk models. The role of the human investor could shift toward higher-level judgment: deciding which research directions are worth pursuing and how to evaluate the results.

With all the excitement around AI, practical concerns remain, including about data privacy, the reliability and traceability of results, the best way to embed rapidly developing technologies in existing workflows and organizations, and the significant cost of deploying large-scale AI systems.

These concerns, along with continued advances in AI, guarantee plentiful conversation in the years to come. Still, Koijen said, he left the gathering convinced that finance is entering a whole new phase.