What elite sport can teach charities about data governance for better AI outcomes
Curia AI | AI Governance Advisory | Insights & Perspectives
I spent a couple of years at MyLife Digital supporting one of the world’s great sporting institutions to understand performance through data, not as an analyst, but close enough to understand what the discipline was really about and what it took to do it well. As a data professional and a lifelong rugby fan, it was a fascinating environment, and one lesson has stayed with me ever since.
The unreliable witness
There is a question I heard one of our senior performance analysts ask coaches early in any new relationship. Not about tactics or squad composition, but this: “In the match you watched last weekend, what percentage of what actually happened do you think you can accurately recall?”
The answers were always revealing. Most coaches would say sixty, seventy, perhaps eighty per cent. They were wrong, consistently and measurably so. The research the team worked with suggested the real figure was closer to thirty to forty per cent, and what coaches did remember was shaped by drama, by recency, and by the confirmation of things they already believed. They over-recalled the decisive moments, the events that fit the story they were already telling. They under-recalled the patterns: the slow erosion of a defensive line, the repeated technical fault that only became visible across fifteen possessions. Nobody was lying. Nobody was incompetent. But human memory under cognitive load is simply not a reliable data collection instrument.
What coaches remembered was shaped by drama, recency, and the confirmation of things they already believed. What they forgot was often more important.
Performance analysis was never about replacing the coach’s expertise. It was about giving that expertise something reliable to work with: consistently defined, carefully captured, longitudinally stored knowledge that compounded in value over time. The analysts maintained a recognition and coding manual that ran to well over a hundred pages. It was not glamorous work. Nobody quoted it in press conferences. But without it, data collected by different analysts across different seasons would have been meaningless, or worse, misleading.
The institutional memory problem
There was also something else I observed: the vast amount of performance knowledge that lived nowhere except in people’s heads and in unstructured files on laptops. GPS records, injury patterns, coaching notes, match observations scribbled in notebooks. When experienced analysts moved on, as they inevitably did, that knowledge moved with them. The game’s institutional memory was fragile and dispersed, not because anyone had been careless, but because no one had ever designed a system to capture it.
I witnessed a specific moment that crystallised this for me. A coaching team was preparing for an opposition they had faced two years earlier. The information they needed existed. It had been observed, analysed, and discussed by the previous analyst. But it had not been systematically retained in a form they could access and interrogate. They were, in practical terms, starting again.
The value was never in the data itself. The value was in a system that captured, structured, and preserved knowledge so that it compounded over time rather than evaporating with each personnel change.
A familiar pattern in an unfamiliar place
Years of working with UK charities have shown me the same challenge playing out in a very different environment. The knowledge that drives great fundraising, why a particular appeal resonated with a particular audience at a particular moment, what the creative hypothesis was behind a campaign, what the team learned from a test and how that shaped the next decision, lives in campaign briefs, in debrief documents, in SharePoint folders, in emails, and most of all in the memory of the fundraiser who ran the campaign.
When that fundraiser moves on, the knowledge moves with them. The next person who runs a similar campaign starts largely from scratch, makes different assumptions, repeats some of the same mistakes, and rediscovers some of the same insights at significant cost. The charity sector has invested enormously in understanding who its supporters are. It has barely begun to invest in understanding why its fundraising works.
This is not a technology failure. It is a data governance failure. And it is one that AI will amplify rather than solve unless the foundations are right.
What AI changes, and what it doesn’t
The conversation around AI can obscure as much as it illuminates. Artificial intelligence does not solve the unstructured data problem; it amplifies it. Poorly structured, inconsistently defined, historically incomplete data has always produced unreliable outputs. What AI changes is the speed and apparent confidence with which those outputs arrive. The governance problem precedes and supersedes the technology question, always.
What AI genuinely changes is the feasibility of capturing and structuring the unstructured. For the first time, it is practical to take the tacit knowledge that lives in campaign briefs, creative rationales, and test-and-learn outcomes and translate it into structured, queryable, longitudinally meaningful data. The analytical capacity now exists to surface patterns across hundreds of campaigns that no individual fundraiser could hold in their head. The institutional memory problem I watched play out in elite sport is, for the first time, genuinely solvable in the charity sector, but only if the data architecture and AI governance framework is right.
Without that foundation, AI tools become sophisticated ways of automating confusion. The charities which will build durable, trustworthy, effective AI capabilities are not the ones who move fastest. They are the ones who build the right foundations first.
A final thought about memory
I think about those coaches often. They were not wrong to trust their instincts. Their instincts were built on decades of accumulated experience and were genuinely valuable. What the data gave them was not a replacement for that expertise, but a foundation that made the expertise more powerful, more reliable, and less dependent on one person’s recollection of one afternoon.
Fundraisers are no different. The experienced fundraiser who knows intuitively that a particular audience responds to a particular kind of emotional narrative has earned that intuition. But when they leave, that intuition leaves too, unless there is a system that has been capturing, structuring, and preserving the evidence that formed it. The system now exists to do this in the charity sector. The only question is whether the sector has the governance maturity and the strategic ambition to use it well.
I have written about this in full in the article linked below, drawing the line between what I observed in elite sport and what I have seen across years of working with UK charities on data strategy, responsible AI, and the trustworthy use of supporter data. I would be very interested to hear whether your organisation has found a way to retain and compound fundraising knowledge over time, or whether you see the same institutional memory problem playing out in your own teams.
What elite sport can teach charities about data and context
The full article explores the governance foundations that allow AI to compound fundraising intelligence rather than amplify its gaps. Download your copy below.
Curia AI helps UK charities and purpose-led organisations build the governance foundations that make AI trustworthy, effective, and strategically valuable.
© 2026 Curia AI Limited · curiaai.co.uk · jcromack@curiaai.co.uk
