What IBM’s AI at Wimbledon can teach Downing Street

IBM’s technology at the Championships is a model of how to deploy AI without the hype. The government’s own AI programme could learn from it.

For two weeks each summer the All England Club becomes one of the most watched technology showcases in the country, though almost nobody thinks of it that way. IBM has been Wimbledon’s technology partner for more than three decades, and the work it does behind the scoreboards has quietly become some of the most ambitious fan-facing AI in British sport.

What that work looks like has changed a great deal. New for this year’s Championships is Match Chat, a generative AI assistant built on IBM’s watsonx platform that answers fan questions during live singles matches, from who has won more break points to how a player tends to hold serve. The system reads a live feed of match data including serve speed, shot direction, rally length and shot type, then turns it into something a casual viewer can follow. Next to it sits an enhanced Likelihood to Win tool that revises each player’s projected win percentage as a match swings.

The early signs are strong. IBM has said engagement with Match Chat is running around 300 per cent higher than the equivalent features it launched the year before. Earlier this year the All England Club was named Sports Organisation of the Year at the Sports Technology Awards for the work.

Two things about the model deserve the attention of anyone running a government department. The first is discipline. IBM and the club run a year-round innovation cycle, sketching ideas in spring, testing them quietly behind the scenes at the US Open and at Wimbledon itself, then deciding in the autumn what ships the following summer. Each feature is measured against whether it actually lifts engagement. Nothing reaches the public until it has been tried in private.

The second is the relationship between machine and human. IBM is explicit that the technology supports its editorial team rather than replacing it. Automation handles the volume and the anomalies so that people can concentrate on judgement. That framing is the opposite of how AI is usually sold to the state.

This is where Whitehall should look up. The government’s AI Opportunities Action Plan and its Humphrey suite of tools for civil servants, Consult, Parlex, Redbox and the rest, are built on a sound instinct. The Consult tool alone is meant to sort consultation responses far faster and more cheaply than officials working by hand. The danger is that government treats AI as an announcement rather than a practice. Wimbledon’s lesson is that credible AI is unglamorous. It rests on narrow use cases, constant testing, honest measurement and a long partnership rather than a one-off procurement.

Departments that want AI to stick could do worse than study a tennis club in SW19. The trophy tends to go to the institution that ships what works and quietly bins what does not.