Craig Hepburn made a point on a recent Unofficial Partner podcast that is worth putting plainly. If every AI company in the world stopped building today, no new models, no new products, no new capability, the gap between what already exists and what most sports organisations are actually doing with it would still take five to ten years to close.
That is not a criticism. It is an observation about where we are. AI adoption in sport is at 82% according to the 2026 Sportradar Global SportsTech Report, which sounds like substantial progress. But adoption is not the same as capability. Most of that 82% covers AI applied to on-field problems: match analysis, injury prediction, recruitment modelling. Clear use cases with clear outcomes. The commercial and operational applications, where the off-field revenue opportunity sits, are much less developed.
The gap Hepburn is pointing at is a knowledge gap. Not a technology gap. The distance between what is possible now, with the tools and models currently available, and what decision-makers inside elite sports organisations believe is possible for their specific commercial operation. Most executives are not underestimating AI in the abstract. They are underestimating what it could do for their specific revenue challenge, right now, with the data and infrastructure they already have.
Why the gap exists
The knowledge gap is primarily an information and translation problem, not a technology problem.
The AI capability that exists today, in commercial applications, in agent frameworks, in connecting proprietary data to intelligent models and building products on top of that connection, has not been communicated clearly to the people who make commercial decisions in elite sport. The people who understand the technology deeply tend to talk in technical terms. The people who make commercial decisions tend to think in revenue lines, partnership structures, and cost bases. The translation between the two has not been made cleanly or consistently.
The result is a pattern that repeats across almost every conversation we have had with Premier League clubs and national stadiums. A commercial director who understands that AI is important, is watching what other organisations are doing, and is being briefed at regular intervals, but has not been given a concrete picture of what AI could do for their specific off-field revenue challenge right now. Not a general AI market briefing. A specific map: here is your current commercial operation, here are the two or three AI applications with the highest return for this venue in the next twelve months, and here is what that is worth.
That map does not require a multi-year transformation programme. It requires someone who understands the available technology well enough to identify the highest-value applications for this specific organisation, and who understands elite sport commercial operations well enough to connect those applications to the revenue lines that actually matter.
What the gap costs in practical terms
The majority of AI investment across elite sport continues to flow toward on-field applications. A smaller share goes to commercial and off-field applications. And a smaller fraction still is structured as genuine infrastructure investment, the kind that creates durable commercial advantage rather than a one-off capability pilot.
The cost of the knowledge gap is not only the investment that is not being made. It is the commercial value that is not being captured. The organisations that have started connecting their venue data to AI commercial tools are already seeing measurable results across hospitality conversion, inbound enquiry volumes, and partnership packaging. The gap between that and where most clubs currently are is not a technology gap. It is a knowledge and operating model gap.
There is also a compounding effect. The organisations that start building AI commercial capability now accumulate data, operational knowledge, and institutional learning that becomes more valuable over time. The clubs that wait a further twelve to eighteen months do not just miss that window of revenue. They start further behind in terms of the data asset that makes AI applications commercially useful in the first place. The gap between early movers and late movers does not stay constant. It grows.
What the applications actually look like
The most useful way to address the knowledge gap is to be specific about what the applications are. Not “AI can improve fan engagement”, which is the general level most commercial directors have already absorbed. Specific use cases, with specific revenue implications for venues running the kind of commercial calendar that Premier League and national stadium clients operate.
Hospitality and events pipeline
AI models trained on a venue’s historical hospitality data can identify which corporate accounts are most likely to convert, at what price point, and at which point in the sales cycle. That is not a complex integration. It is a relatively straightforward application of a language model to structured commercial data, combined with a workflow change in how the hospitality team manages its pipeline. For a venue running 200-plus non-matchday events per year, the revenue impact of improving conversion on even a modest proportion of that pipeline is material.
AI visibility for venue discovery
When a corporate buyer asks ChatGPT, Gemini, or Perplexity for event venues in London for 150 people, which venues appear in the answer? Most commercial directors have not tested this. The answer is often surprising. Venues that have invested in the right structured data and content formats appear consistently. Venues that have not are effectively invisible to AI-mediated discovery. This is a specific, fixable, measurable problem, and fixing it has a direct impact on inbound enquiry volume. Our resources page covers the detail of how this works in practice.
Dynamic commercial packaging
AI models can process partner and sponsor data, content performance signals, and fan behaviour data to identify commercial packaging combinations that would not surface through conventional analysis. This is particularly relevant for national stadiums and arenas with complex, multi-sport commercial calendars where the interaction between different revenue lines is hard to model manually.
None of these require a transformation programme. Each requires a combination of the right data access, the right model configuration, and the right workflow change in the commercial team. The constraint is not the technology. It is knowing where to start and what to do first.
The role of evaluation in closing the gap
One of the most underappreciated shifts in how AI commercial work operates is the change in where the heavy work sits. It has moved from building to evaluating.
In the old model, the constraint was build time. If you wanted to test whether a product or application was valuable, you had to build it first, which took months and significant cost. You deployed it, and then found out whether the original hypothesis was right. If it was not, you started again.
The constraint has moved from building to evaluating: determining whether what has been built is actually delivering the right commercial outcome, and iterating on that basis. This is different work. It requires different skills and a different mindset. It is not about producing more output faster. It is about asking better questions of the output you can now generate quickly and cheaply.
Hepburn described this well: the bulk of the work he sees right now is not building and developing. It is evals, validating outcomes, researching whether the platform delivers the right value, making sure the information the AI is using is accurate. Whereas before, once you had put all the effort into building and deploying something, it was a substantial piece of work to change it. Now you can iterate continuously. The value has moved upstream to the quality of the question and the rigour of the evaluation.
For commercial teams at elite sports organisations, this means the most valuable external capability is no longer someone who can produce things. It is someone who can help the team figure out which things are worth producing, test them quickly against real commercial data, and make confident decisions about what to build towards at scale.
Closing the gap in practice: the sequence that works
The clubs that close the knowledge gap fastest are not necessarily the ones with the largest technology budgets. They are the ones that make a deliberate decision to understand specifically what AI could do for their commercial operation, with their current data and infrastructure, in the next six to twelve months.
The sequence that works in practice has three stages.
1. Get a clear picture of the current state. Where does this venue sit relative to AI-capable peers, what is the gap in specific commercial terms, and what are the priority applications ranked by revenue impact and implementation complexity. This is a diagnostic process, not a general briefing, and it produces a specific output that a commercial director can act on. Earl’s AI Visibility Diagnostic is built around this: a 4-6 week, fixed-fee process that produces that picture clearly and without the noise.
2. Run the priority applications. Connect the right data to the right models, build the workflow changes in the commercial team, and establish the evaluation framework that allows the team to measure what is working and what is not. This is where the operating model change happens. It is also where the institutional knowledge builds, which is part of what makes early movers harder to catch later.
3. Build the infrastructure layer. The data platform, the APIs, the agent framework that allows the club to keep building as the AI landscape evolves, rather than starting from scratch with each new capability. This is the work that converts a good first-year result into a durable commercial advantage.
The knowledge gap is large, and it is widening. But it is closable, and the tools required to close it exist right now. Hepburn’s observation works in both directions. The gap between where most clubs are and where they could be is significant. But everything required to close it is already available. The capability is not coming. It is here.
The question is not whether your commercial operation will eventually need to engage with this seriously. It is whether you start building the understanding and the infrastructure now, or whether you find yourself, in two or three years, in the position of catching up to organisations that started earlier and have compounded their advantage in the meantime.
That is a commercial planning question, not a technology question. And it is worth answering clearly.