Can AI be the great equaliser in e-FX?

FX market-makers see real benefits for agentic AI in code generation and data analysis

Key points Summary Q&A

• Claude Opus 4.6, released in February, has raised hopes that AI can narrow the capability gap in spot FX market‑making.
• Regional banks say agentic AI can rapidly generate high‑quality code, speeding new features and model development without large tech teams.
• A senior regional e‑FX trader says Claude’s code is almost production-ready, and the tools also speed data analysis and continuous signal scanning.
• Tier‑one banks are testing agentic AI too and may use it to analyse liquidity pools and refine pricing, so AI could both close and widen gaps.

These summaries are generated by an LLM, but edited by humans.

Software developers are not the only ones excited by the release of the Claude Opus 4.6 agentic artificial intelligence model in February.

Over the years, a significant gulf in trading capabilities has been apparent in the world of electronic foreign exchange market-making. Constant pressure on margins has made it difficult for any bank, let alone a regional one, to invest the resources needed to compete effectively with the largest players. 

The recent advances in AI, however, have the potential to rapidly narrow this capability gap. Now, regional banks are hopeful that the technology can put them on an equal footing with their tier-one peers in the cut-throat world of spot FX liquidity provision.

How could this happen? 

It is important to distinguish what AI has already been used for. All FX execution algorithms and e-FX pricing engines across most banks have at their core some form of machine learning, either built into their order routers to manage distribution of flows or to dynamically update pricing. 

A senior e-FX trader at one regional bank believes the quality of the code being generated by Claude has reached a level where it can almost be put straight into production

But some regional e-FX market-makers see the real benefits of agentic AI on the code generation side. They lack the large tech teams of their bulge-bracket rivals, so are hopeful that using this technology will help them deliver innovation in a fraction of the time it has taken up until now. 

Instead of asking a team of quants to spend weeks writing code for a new feature within their execution algorithms or to price a certain instrument in a new way, for instance, they could just ask the model to generate it within minutes. It could also be used to help analyse and validate code to enhance decision-making for electronic trading teams. With spot FX sitting outside the regulatory perimeter, some believe a little more leeway exists for these teams to use the technology in areas such as model development.

A senior e-FX trader at one regional bank believes the quality of the code being generated by Claude has reached a level where it can almost be put straight into production. And it is not only coding where the tools are coming in handy – the tech can be used for faster data analysis that might previously have required specialist teams, while autonomous AI agents can constantly scour data points for important signals or information.

Related

How FX pricing is adapting to Trumpian markets 

Dealers are tying pricing engines to new signals in effort to cope with out-of-the-blue moves

Could this become the great equaliser, enabling regional banks – now equipped with the new AI tools – to compete with the bigger players?

Agentic AI is also available to tier-one banks, who are carrying out their own experiments. Yes, it’s taken a bit longer owing to lengthier checks and controls required for updates, but it is understood that progress is rapidly being made to utilise the technology for coding and beyond.  

For instance, an e-trading desk could use it to pull in data across all the liquidity pools they operate and better analyse market structure more quickly. It could help provide rapid insights into where they may be suffering from adverse selection and where they need to refine their pricing, or create customisable trading strategies for specific clients based on the real-time data feeds from the platforms.

So, while the new tools may allow regional banks to play catch-up in some areas, they could also widen the gap with the larger players in other ways.

Editing by Lukas Becker