From Playbooks to Protocols: Learning from High-Reliability Industries to Govern AI in Sport

<!-- Article -->
<main class="article-wrap">

<!-- Hook -->
<div class="hook">
<p><strong>Everyone's asking how fast they can adopt AI.</strong></p>
<p>Nobody's asking what happens when it goes wrong.</p>
<p>That's not a technology problem. <strong>That's a governance problem.</strong></p>
<p>Aviation doesn't wait for a crash to review its systems. Medicine doesn't wait for a patient to die before questioning the protocol. These industries build oversight into the culture long before anything goes wrong.</p>
<p>So why does sport wait for a governance failure before asking: <em>are we actually ready for AI?</em></p>
</div>

<!-- Section 1 -->
<h2>The Seduction of Speed</h2>

<p>Let's be honest. The pressure is real.</p>
<p>Boards are hearing that AI will transform performance analysis. That it will predict injuries before they happen. That it will unlock insights human coaches simply can't see. And yes — in the right conditions, with the right data and the right oversight — some of that is true.</p>
<p>But <strong>"some of that is true"</strong> is doing a lot of heavy lifting right now.</p>
<p>Because AI doesn't just amplify your strengths. It amplifies your weaknesses too. And if your governance structure isn't built to catch errors — to sense when something's drifting, to course-correct before it becomes a crisis — then the speed of your AI adoption is exactly the speed at which your vulnerabilities will surface.</p>
<p>That's the part nobody's talking about at the board table. Yet.</p>

<!-- Section 2 -->
<h2>What High-Reliability Organisations Already Know</h2>

<p>There's a field of study called <strong>High-Reliability Organisation theory</strong> — HRO for short. It emerged from research into industries where failure is catastrophic and largely irreversible: commercial aviation, nuclear energy, emergency medicine, naval operations.</p>
<p>In these environments, you don't get to say "we'll learn from that mistake" when the mistake kills people or crashes a system thousands depend on. So over decades, researchers <em>Karl Weick</em> and <em>Kathleen Sutcliffe</em> identified the specific organisational behaviours that allow complex systems to stay reliable under pressure.&nbsp;¹</p>
<p>They found five consistent principles:</p>

<ol class="principles">
<li><strong>Preoccupation with failure.</strong> HROs treat near-misses as gifts — early warnings that something in the system is weak. They don't celebrate getting away with it. They investigate why it almost went wrong.</li>
<li><strong>Reluctance to simplify.</strong> These organisations resist reducing complexity into a tidy story. They know oversimplification is where blind spots breed.</li>
<li><strong>Sensitivity to operations.</strong> Leaders stay connected to what's actually happening on the ground — not just what the dashboard reports.</li>
<li><strong>Commitment to resilience.</strong> The system is designed to absorb disruption and recover — not just to optimise performance when everything's going well.</li>
<li><strong>Deference to expertise.</strong> Decision-making authority shifts to whoever holds the most relevant knowledge in that moment — not just whoever holds the title.</li>
</ol>

<p>Now. Read that list again.</p>
<p>And ask yourself honestly: how many of those principles describe how your board governs AI right now?</p>

<!-- Section 3 -->
<h2>Why Sport Governance Isn't Built for This — Yet</h2>

<p>Sport organisations were built for a different kind of complexity. The traditional governance model — board oversight, annual reporting cycles, strategic planning in four-year Olympic blocks — made sense when the pace of change was manageable.</p>
<p>It doesn't make the same sense anymore.</p>
<p>The <strong>NIST AI Risk Management Framework</strong> — one of the most robust voluntary standards for AI governance internationally — identifies four core functions: <em>Govern, Map, Measure, and Manage.</em>&nbsp;² The parallel with HRO thinking is striking. The NIST framework, developed independently, arrives at remarkably similar conclusions: don't assume AI will always work correctly — build your oversight structures as if it won't.</p>
<p>But here's the gap. Most sport organisations haven't mapped their AI risk exposure. They haven't defined who is accountable when an algorithm makes a call that harms an athlete's career, misidentifies talent, or produces a biased output. They haven't created the channels that would allow a staff member to raise a concern about a flawed model before it cascades.</p>
<p>Based on practitioner evidence and the governance gaps documented in the field, a clear pattern is emerging: <strong>boards claim oversight of AI initiatives, but operational governance is often thin</strong>. Steering committees exist on paper. Ethics principles appear in strategy documents. But the lived reality — the governance that actually shapes decisions in real time — is largely absent.&nbsp;³</p>
<p>And it's exactly what HRO theory warns about: the gap between <em>stated</em> governance and <em>lived</em> governance is where failures incubate.</p>

<!-- Section 4 -->
<h2>The Athlete Is Not a Data Point</h2>

<p>And the people who bear the consequences of that gap most directly? They're not in the boardroom.</p>
<p>Here's something that gets lost in conversations about AI adoption: <strong>the people most affected by algorithmic decisions in sport are athletes.</strong></p>
<p>Injury prediction models influence training loads. Talent identification systems shape selection decisions. Performance analytics influence contract negotiations. Biometric data is collected, stored, and analysed — often without athletes having meaningful input into how it's used or what safeguards exist.</p>
<p>Human rights frameworks are increasingly clear on this. The <em>Council of Europe's Framework Convention on Artificial Intelligence</em> and the <em>OECD AI Principles</em> both emphasise that AI systems affecting individuals must include accountability mechanisms, transparency, and meaningful avenues for redress.&nbsp;⁴&nbsp;⁵</p>
<p>But in sport, we haven't caught up.</p>
<p>The evidence from human rights and athlete welfare frameworks points consistently in one direction: athletes are still largely treated as data sources rather than governance partners. Their voice — when it exists at all — tends to come through athlete commissions that sit adjacent to governance rather than embedded within it.&nbsp;⁶</p>
<p>A truly <strong>athlete-first</strong> approach to AI governance doesn't just mean protecting athletes from harm. It means involving them in the design of the systems that shape their careers. It means creating formal mechanisms for athletes to raise concerns about AI-driven decisions. It means building <em>their</em> expertise into the governance structure — not as a box to tick, but as a source of intelligence the board genuinely can't get elsewhere.</p>

<div class="pull-quote">
<p>Great systems elevate people. Poor systems constrain them. That includes athletes.</p>
</div>

<!-- Section 5 -->
<h2>From Playbooks to Protocols: What This Actually Looks Like</h2>

<p>So what does HRO-informed AI governance look like in practice?</p>
<p>It doesn't mean turning your national federation into a nuclear plant. The stakes are different. The operational context is different. The regulatory regime is different. <strong>Analogies must be adapted, not transplanted.</strong></p>
<p>But the underlying logic transfers directly.</p>

<div class="step">
<div class="step-header">
<span class="step-num">Step 1</span>
<span class="step-title">Build a preoccupation with failure into your review cycles</span>
</div>
<div class="step-body">
<p>Don't just report on what went right. Build formal structures — in board reporting, in HP systems reviews, in post-Games debriefs — that specifically interrogate where AI-assisted decisions may have drifted, where data quality was questionable, where model outputs contradicted expert judgment.</p>
<p>Treat every near-miss as a signal, not an anomaly.</p>
</div>
</div>

<div class="step">
<div class="step-header">
<span class="step-num">Step 2</span>
<span class="step-title">Resist the simplification trap</span>
</div>
<div class="step-body">
<p>AI dashboards are seductive. They convert complexity into clean visualisations and decisive-sounding recommendations. That's their value — and their danger. Build deliberate friction into how your board reviews AI outputs: require that model recommendations are challenged by domain experts before action is taken.</p>
<p>Require that the <em>limitations</em> of any model are surfaced alongside its outputs — not buried in a technical appendix.</p>
</div>
</div>

<div class="step">
<div class="step-header">
<span class="step-num">Step 3</span>
<span class="step-title">Define accountability before you need it</span>
</div>
<div class="step-body">
<p>One of the most important questions any sport board should be able to answer is: <em>if an AI-driven decision causes harm to an athlete — who is responsible?</em> Not in theory. Not in the strategy document. In practice, right now, with the systems you currently have deployed.</p>
<p>If that question takes more than thirty seconds to answer, you have a governance gap.</p>
</div>
</div>

<div class="step">
<div class="step-header">
<span class="step-num">Step 4</span>
<span class="step-title">Create expert deference pathways</span>
</div>
<div class="step-body">
<p>In high-stakes moments, the most reliable organisations shift authority toward whoever holds the most relevant expertise — not the most senior title. Build this into your AI governance: ensure that technical experts, data scientists, and — critically — the athletes whose performance is being modelled have formal pathways to influence decisions.</p>
<p>Not just advisory roles they can be quietly ignored in.</p>
</div>
</div>

<div class="step">
<div class="step-header">
<span class="step-num">Step 5</span>
<span class="step-title">Build resilience, not just optimisation</span>
</div>
<div class="step-body">
<p>HP sport has a tendency to design systems that perform brilliantly when everything is going well. HRO theory asks a different question: <em>how does this system perform when something unexpected happens?</em></p>
<p>Design your AI governance not just for peak performance — design it to recover gracefully from model errors, data quality failures, and outputs nobody predicted.</p>
</div>
</div>

<!-- Section 6 -->
<h2>The Funding Reality Nobody Talks About</h2>

<p>One more thing. And this one's for everyone who just thought: <em>"That sounds great, but we don't have the budget for it."</em></p>
<p>The case studies that dominate the AI-in-sport conversation tend to feature elite clubs with nine-figure budgets, Olympic programmes with dedicated data science teams, and leagues with enterprise-level infrastructure.</p>
<p>That's not most sport organisations — many operate technology budgets that could be covered by a single athlete's annual funding grant.</p>
<p>Research from Sport New Zealand confirms that many grassroots and mid-tier organisations are significantly behind in AI readiness, not because of a lack of will, but because of real resource constraints.&nbsp;⁷ And here's the uncomfortable truth: <strong>resource-constrained organisations are precisely the ones most exposed to AI governance failures</strong>, because they have the least capacity to build the oversight structures that responsible adoption requires.</p>
<p>An AI tool that costs a federation relatively little to license may create data privacy exposure, liability risk, or athlete welfare consequences that the organisation has no governance infrastructure to manage.</p>
<p>In my work with national federations and smaller NGBs, the pattern is consistent: the tool gets adopted fast, the governance comes later. Sometimes much later.</p>
<p>The responsible approach isn't to wait until the budget allows perfect governance. It's to calibrate AI adoption to your <em>current</em> governance capacity — and to treat governance investment as a prerequisite for AI investment, not an afterthought to it.</p>

<!-- Closing -->
<h2>The Question Worth Asking</h2>

<p>This is not an argument against AI in sport.</p>
<p>It's an argument for governing it properly.</p>
<p>The organisations that will get this right aren't the ones that move the fastest. They're the ones that build the governance infrastructure to move with <em>confidence</em> — that create the protocols, not just the playbooks.</p>
<p>Aviation doesn't wait for a crash to review its systems. It builds the review into the culture long before anything goes wrong. That's not caution. That's intelligence.</p>
<p>That's the shift sport needs to make.</p>
<p class="closing-bold">The question isn't whether AI belongs in your high-performance system. It does.</p>
<p class="closing-bold">The question is: is your governance built for what comes with it?</p>

<!-- CTA -->
<div class="cta-block">
<h3>Find Out Where Your Organisation Stands</h3>
<p>Your next Olympic cycle starts with this question. The HP Systems Readiness Scorecard gives you a personalised assessment of your organisation's current position — and a clear picture of what to strengthen before your next cycle begins.</p>
<a href="https://glenviewsports.co/scorecard" class="cta-btn">Take the HP Systems Readiness Scorecard &rarr;</a>
</div>

<!-- References -->
<div class="references">
<h2>References</h2>
<ol>
<li>Weick, K. E., &amp; Sutcliffe, K. M. (2015). <em>Managing the Unexpected: Sustained Performance in a Complex World</em> (3rd ed.). Jossey-Bass.</li>
<li>National Institute of Standards and Technology. (2023). <em>Artificial Intelligence Risk Management Framework (AI RMF 1.0).</em> <a href="https://doi.org/10.6028/NIST.AI.100-1" target="_blank">https://doi.org/10.6028/NIST.AI.100-1</a></li>
<li>Sports Governance Academy. (2025). AI Governance — Why does it matter? Databloom Partners / Chartered Governance Institute. <a href="https://www.sportsgovernanceacademy.org.uk/resources/blog/ai-governance-why-does-it-matter/" target="_blank">sportsgovernanceacademy.org.uk</a></li>
<li>Council of Europe. (2024). Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law. <a href="https://www.coe.int/en/web/artificial-intelligence/the-council-of-europe-framework-convention-on-ai" target="_blank">coe.int</a></li>
<li>OECD. (2024). OECD Principles on Artificial Intelligence. <a href="https://oecd.ai/en/ai-principles" target="_blank">oecd.ai/en/ai-principles</a></li>
<li>Article One Advisors. Artificial Intelligence &amp; Human Rights in Sport. <a href="https://articleoneadvisors.com/artificial-intelligence-human-rights-in-sport/" target="_blank">articleoneadvisors.com</a></li>
<li>Sport New Zealand. (2025). <em>AI in Sport and Recreation 2025 Report.</em> <a href="https://sportnz.org.nz/media/wmbdkaxx/ai-in-sport-and-recreation-2025-report.pdf" target="_blank">sportnz.org.nz</a></li>
<li>OSHwiki — European Agency for Safety and Health at Work. High Reliability Organisations. <a href="https://oshwiki.osha.europa.eu/en/themes/high-reliability-organizations" target="_blank">oshwiki.osha.europa.eu</a></li>
<li>Sport Information Resource Centre. (2024). Beyond the Buzzwords: A Practical Guide to AI for Sport Leaders. <a href="https://sirc.ca/articles/beyond-the-buzzwords-a-practical-guide-to-ai-for-sport-leaders/" target="_blank">sirc.ca</a></li>
</ol>
<p style="font-size:0.78rem; color: #888; margin-top: 16px; font-style: italic;">All URLs reference publicly accessible, non-commercial resources from government agencies, academic institutions, and professional bodies.</p>
</div>

</main>