Applying Howard Marks' AI Framework to Sports Participation | April 2026
In February 2026, legendary investor Howard Marks published a memo that rattled Wall Street. Not because of a market call — but because of a simple three-level framework for understanding artificial intelligence. Clear enough to cut through the noise that has surrounded AI discourse for years.
Marks was writing for investors. But his framework belongs to anyone trying to navigate what AI actually means for their sector. For those of us working in sport — governing bodies, state sporting organisations, clubs and community administrators — it offers something rare: a way to think about AI that is honest about where the opportunity lies, and where the hype ends. Practical and executable. Exactly what sport needs.
Participation is the foundation of sport. Lifelong habits form here. Community bonds are built here. The pipeline to elite performance begins here. It is also chronically under-resourced. Governing bodies are asked to grow participation with thin budgets, ageing volunteer bases, and increasingly fragmented attention from the communities they serve.
If AI can genuinely help, understanding how — and at which level — is worth getting right.
"Level 3 is autonomous agents... This is labour replacement at the task level." — Howard Marks, AI Hurtles Ahead, 2026
Level 1: AI as Conversation — The Research Assistant
What Marks Says
In Marks' framework, Level 1 is the AI most of us have already encountered. You ask a question; you receive an answer. It saves research time. It does not execute on your behalf. Its value is real but bounded.
What This Looks Like in Grassroots Sport
For a sporting organisation, Level 1 AI is deployable today — without significant investment or technical expertise. A community club secretary asks an AI tool to draft a funding submission. A volunteer coach asks it to generate a junior training session plan for nine-year-olds. A participation officer asks it to write social media posts for a Come and Try day.
These are real time savings. For volunteers who are already stretched, reducing the cognitive load of administrative writing is meaningful. A script for a first-timer event, a grant application outline, a parent information letter — tasks that once took hours can now take minutes.
The Honest Nuance
Level 1 AI does not know your community. It does not know that your local club has a specific barrier around female participation after 6pm, or that your junior numbers dropped because a rival code started a program at the same school. The output is generic unless the inputs are specific. Sports organisations need to develop the skill of prompting well — giving AI the context it needs to produce something genuinely useful rather than blandly competent.
The risk at Level 1 is complacency: believing that because AI can write the newsletter, the thinking is done. AI at this level is a tool that amplifies human intent. Weak intent produces weak output.
Level 2: AI as Tool — The Analyst and Executor
What Marks Says
Level 2 is where AI becomes meaningfully more powerful. You instruct it to find information, analyse it, and perform tasks. The economic value is larger because it saves execution time, not thinking time alone. It is still bounded — it does what you direct it to do.
What This Looks Like in Grassroots Sport
Level 2 is where sports administrators should be most actively experimenting in 2026. It can analyse registration data to identify which postcodes are underperforming relative to demographic benchmarks. It can cross-reference attendance records with school term dates, local events, and weather to identify patterns in drop-off. It can monitor social media sentiment around a club to flag early signs of community concern before they become a crisis.
Consider the participation challenge every governing body faces: attracting and retaining participants. Level 2 AI can be configured to analyse registration and attendance data, identify at-risk participants based on engagement patterns, and generate personalised re-engagement communications — all within a system the organisation manages.
The Honest Nuance
Level 2 AI requires data. Grassroots sport has a data problem. Many clubs still manage registrations through spreadsheets, WhatsApp groups, and institutional memory. Others use a platform, but extracting relevant historical data remains a challenge. Before Level 2 AI can deliver its promise, organisations need to invest in basic data infrastructure: consistent registration systems, digital attendance tracking, centralised communication records.
There is also an equity question worth naming. Level 2 AI will surface insights about the communities already engaged in your data systems. The hardest participation challenge — reaching communities who have never registered, never attended, never engaged digitally — will not be solved by AI alone. Human relationships, community trust, and culturally competent outreach remain irreplaceable. AI at this level amplifies those capabilities; it does not substitute for them.
Level 3: AI as Agent — The Autonomous Operator
What Marks Says
Level 3 is where Marks stops being measured and starts sounding almost alarmed. You give AI a goal and parameters, and it does the work, checks it, and delivers a finished product — without step-by-step instruction. Marks is direct: this is labour replacement at the task level. As of early 2026, we have arrived here. AI models can now run multi-step projects, manage their own sub-tasks, and deliver completed outputs with minimal human direction.
What This Looks Like in Grassroots Sport
Level 3 AI in a sports participation context is still emerging, but the trajectory is clear. A participation officer begins Monday morning by reviewing a report an AI agent compiled overnight: participation numbers by club, gender, and age bracket; a draft response to the three most pressing parent enquiries received over the weekend; a revised schedule for the upcoming development day based on venue availability that changed Friday; a shortlist of funding opportunities with draft expressions of interest for each.
The AI agent did not wait to be asked. It was given a goal — support participation growth and operational efficiency — and it worked. The participation officer's role shifts from doing to directing, reviewing, and connecting with community in ways that require human presence.
The Honest Nuance
Level 3 demands a conversation that sports organisations should have an eye on — without alarm. In a resource-constrained environment, this could be a profound opportunity rather than an existential threat.
Governing bodies need to be explicit about intent. Using Level 3 AI to expand participation reach — to do more with the same people, not fewer people doing the same — is a fundamentally different proposition to using it to cut headcount. The organisations that thrive will be those who set those goals clearly and early.
Where Should Governing Bodies Start?
Marks' three levels are not a roadmap that demands you start at Level 1 and work upward. They are a diagnostic tool. The right starting point depends on your organisation's current capabilities, data maturity, and strategic priorities.
A governing body with strong data infrastructure and clear participation goals can start experimenting with Level 2 now. A club network dominated by time-poor volunteers will get more immediate value from Level 1 tools. The framework gives you a language for assessing where your organisation sits — and an honest picture of what each level requires before it delivers.
The Bottom Line
Howard Marks wrote his memo because he believes AI is real, consequential, and the worst response is to either dismiss it or assume it will solve everything. That measured view applies exactly to sport.
AI will not fix a participation decline rooted in cost barriers, transport disadvantage, or cultural disengagement. It will not replace the club volunteer who remembers every player's name and makes the new kid feel welcome. It will not substitute for the trust a local coach builds over years in a community.
What it can do is free those people to focus on the work they are irreplaceable at. Surface insights no human analyst had the time to find. Handle the administrative weight that burns out the volunteers sport cannot afford to lose. And at Level 3, operate proactively — working toward the goals we set, without waiting to be asked.
The question for governing bodies is whether AI will be used to grow the sport.
Marks ends his memo with prudence. The technology is real. The value is real. The uncertainty is real too. The organisations that benefit most from AI will not be those who adopt it earliest, but those who adopt it most clearly — with a precise sense of what they are trying to achieve and for whom.
In grassroots sport, that means keeping participation at the centre. AI is a means. More people playing, for longer, in environments where they feel they belong — that is the end game.
This piece draws on Howard Marks' February 2026 investment memo 'AI Hurtles Ahead,' published by Oaktree Capital Management.
