It is the most exciting development in capital markets for many, but attendees at FILS 2026 in Boston may find that what is being shared in panel discussions about artificial intelligence (AI) may represent the more conservative end of what is actually being built.
The most consistent theme across every panel preparatory call where AI was discussed, is that firms are actively watching and experimenting with AI, but that live deployment in core investment and trading workflows remains limited. One attendee put it most plainly: “The honest answer is not yet,” and said his firm is “at the infant stages” and still questioning AI’s role internally.
Another attendee has already observed that AI tools used at firms like Google and Facebook have been available at for years, yet productivity at those organisations has increased “by probably 10% at best” — cautioning against assuming a step-change in output.
Bank-owned investment firms – and banks themselves – are heavily regulated which limiting both their use and public discussion of AI. Nevertheless, even some speakers from independent asset managers have been pulled back from speaking at this year’s event purely because AI discussion was perceived to create risk, by a combination of press and compliance teams.
Yet behind the scenes, some firms say their use is extensive, innovative and inventive, making radical changes to the way that they operate today.
The first use case bridges the shortfall in IT investment found at some buy-side firms. AI is providing a single interface for querying information, using non-technical natural language prompts, accelerating research and data processing workflows that previously consumed significant analyst and trader time.
Agents that ingest internal data, processes and policies can answer questions or build broker assessments in eight minutes.
AI’s role in helping write code and conduct research faster, is enabling a ‘democratisation’ of research capability, so that firms with limited quant resource can ask better questions of large datasets. In some cases, AI is acting like a data scientist, pointing simple models at large, fixed income datasets and extracting signals traders previously lacked the resources to identify.
The capacity to systematise investment processes can be better supported by AI, as it can unify understanding between research, portfolio management, and trading, who are each focused on producing alpha through different ways e.g. signal generation, risk management and portfolio construction, and trade execution.
All of this is clearly supervised, supporting activity and systematic signal outputs need to be used carefully with human judgement. Increased efficiency does not equate to greater accuracy.
A second use case is supporting order management. AI is well-suited, some argue to analysing market conditions, securities’ characteristics, historical trading patterns and available data to guide when and how liquidity should be pursued. The role of AI in this vision is as an overlay, directing protocol selection, rather than replacing the execution process itself.
By providing recommendations of protocol selection this reduces cognitive load on the desk and enables lower-touch execution. However, this remains aspirational to date.
A third case is the parsing of unstructured data such as Instant Bloomberg (IB) messages and dealer axes to make this information more actionable. By overlaying parsed message and axe data with holdings information, conviction changes and trade history desks are trying to provide something usable at-trade.
For all of these, siloed data is often cited as the main constraint, for example, parsed messages are only as useful as the firm’s ability to combine them with holdings and conviction data.
Hearing this discussed on stage is being hampered by the concerns of risk and publicity averse functions, however both on and off stage this will be a major theme this year.
©Markets Media Europe 2025
