AI agents in high-performance football: where should we actually start?

For years, I wanted the knowledge I needed in high-performance football to be organised, connected and available when a decision had to be made. I never really managed to build that system. The work moved quickly, I was carrying responsibilities across several roles, and there was rarely enough time to stop, capture what I knew and make it accessible when I needed it.

That experience is why the current conversation about AI agents feels so close to the work. The more we talk about agents, the clearer the basic requirement becomes: people and agents need a shared, structured base of knowledge and context.

I knew what I needed. I couldn't make space to organise it.

Across my work in high-performance football, I moved between physical preparation, strength work, return-to-play and football coaching. When we were dealing with physical problems, I often had to keep several perspectives in view at once, including questions connected to physiotherapy, nutrition and psychology.

I wanted my decisions to draw on everything relevant: what I had studied at university, what practical work had taught me, and scientific evidence as it evolved. But the day rarely paused for me to capture, process and organise all of it. I never really managed to build the clear, accessible system I wanted.

Some knowledge stayed in my own experience; some was difficult to find when I needed it. The gap between knowing that useful information existed and being able to use it under pressure was frustrating. At times, it was stressful. Decisions came quickly, while organising knowledge required time I rarely had.

AI makes the need harder to ignore

An AI agent needs a defined role, a specific task, clear boundaries and the right context, just like a colleague expected to contribute to a team. That context has to come from knowledge people can find, connect to the workflow and keep current.

If the relevant knowledge lives only in people's heads or is hard to retrieve, it is difficult to give an agent dependable instructions. It is also difficult for staff to understand, review and take responsibility for what the agent produces.

For me, that makes knowledge management a practical obligation. If we want agents to support high-performance work, we have to organise and govern the knowledge they rely on, alongside the people who use it. Professionals remain responsible for decisions; a clear system makes the knowledge available and the agent's role understandable.

The agent is the visible part. The knowledge system gives it direction.

What we are building at SportSync

At SportSync, we are working on this with clients across four connected layers:

• Mapping processes and workflows: understanding how work really happens, how decisions move, where handovers and friction appear, and what knowledge is still tacit.
• Building a structured knowledge base: organising information and its relationships so it reflects the work and can be found and used.
• Creating governance: defining policies, protocols and working rules. Skills add more specific instructions that guide how people and agents carry out particular tasks.
• Developing automations and agents: giving them a defined role in the workflow, with the context, instructions and boundaries they need.

These layers inform one another. Mapping a workflow can reveal knowledge that needs to be made explicit. Structuring that knowledge can expose a missing relationship or an unclear process. Protocols and skills make expectations clearer; work with automations and agents can then show us where the map or the knowledge base needs to improve.

SportSync's methodology is grounded in our experience and knowledge from high-performance football, and shaped by the real problems we work through with clients. We keep refining it through that day-to-day work: mapping workflows, organising knowledge, and defining the context people and agents need.

The work behind the agent

The starting point, for me, is the real work: where the knowledge sits, what decisions depend on it, how the workflow operates, and what has to be made explicit so people and agents can use it.

That is the work we are doing with clients at SportSync now. I’ll keep sharing what we learn from this work. If you’re working through similar questions in your club or performance environment, I’d be happy to stay in touch.