{"id":36156,"date":"2026-09-17T08:00:00","date_gmt":"2026-09-17T07:00:00","guid":{"rendered":"https:\/\/leadersinsport.com\/performance-institute\/?post_type=article&p=36156"},"modified":"2026-09-17T11:20:14","modified_gmt":"2026-09-17T10:20:14","slug":"ai-is-not-the-injury-oracle-you-were-promised-but-it-can-change-what-injuries-cost","status":"publish","type":"article","link":"https:\/\/leadersinsport.com\/performance-institute\/articles\/ai-is-not-the-injury-oracle-you-were-promised-but-it-can-change-what-injuries-cost\/","title":{"rendered":"AI Is Not the Injury Oracle you Were Promised. But it Can Change What Injuries Cost"},"content":{"rendered":"\n\n
\n\t
\n\n\t\t
\n\t\t\t
\n\n\t\t\t\t
\n \n\t\t\t\t\t
\n\t\t\t\t\t\t

\n 17 Sep 2026 <\/p>\n \t\t\t\t\t\t\n\t\t\t\t\t\t\t<\/span>Articles<\/a>\n\n\t\t\t\t\t\t

AI Is Not the Injury Oracle you Were Promised. But it Can Change What Injuries Cost<\/h1>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\n \n\t\t\t<\/div>\n\t\t<\/div>\n\n\t<\/div>\n<\/div>\n
\n
\n
\n
Category<\/div>\n Open Access<\/a> <\/div>\n
\n
Share<\/div>\n Facebook<\/a>\n Twitter<\/a>\n Email<\/a>\n Copy Link<\/a>\n
https:\/\/leadersinsport.com\/performance-institute\/articles\/ai-is-not-the-injury-oracle-you-were-promised-but-it-can-change-what-injuries-cost\/<\/div>\n <\/div>\n <\/div>\n <\/section>\n\n\n\n\n
\n
\n
\n
\n

David Clancy of The NXT Level Group and Ivan Peric of Peric & Partners discuss how artificial intelligence can help your performance team to see more clearly, remember more accurately \u2013 and learn more quickly.<\/p>\n \n \n \n

\n

By David Clancy and Ivan Peric<\/h6>\n

The morning<\/strong><\/p>\n

At 8:20 on a Monday morning, the sports medicine and performance team gathers around a table. The physio updates the group on a player. The club doctor adds context from the recent scan. The lead S&C coach describe yesterday\u2019s exposure. The sports scientist flags a change in high-speed decelerations. The performance nutritionist has noticed a fuelling issue. The Manager wants a simple answer: Will he train? And when will he be ready to play again?<\/p>\n

There is no shortage of expertise in the room. There is no shortage of data either. What is missing is memory. What did the department do the last time this player had this problem? What was expected? What happened? Which decision changed the trajectory? And, with the benefit of hindsight, was it the right one?<\/p>\n

Some of that sits in the digital records. Some sits in disconnected files. Much of it lives in the physiotherapist\u2019s memory, or in the conversation that happened beside the treatment table three months ago. Eleven minutes later, a decision is made. By the afternoon, parts of the rationale have already begun to disappear.<\/p>\n

That is the real opportunity for AI in elite sport. Not to replace the people around the table. Not to produce an all-knowing injury-risk score. Not to make the return-to-play decision. It is to help the team behind the team see more clearly, remember more accurately \u2013 and learn more quickly.<\/p>\n

The bill is \u20ac3.45 billion, and technology has not moved it<\/strong><\/p>\n

Howden\u2019s Men\u2019s European Football Injury Index counted 22,596 injuries across Europe\u2019s top five leagues over five seasons. The estimated wage cost of players being unavailable was \u20ac3.45 billion. The injury rate has barely shifted. One injury every 625 mins. The financial bill has not stood still; it increased by 29% across that five-year period (Exhibit 1).<\/p>\n


\nThe injuries are not distributed evenly. Clubs carrying the heaviest fixture demands tend to record more of them. Four clubs sat above the Premier League injury average in every one of the five seasons. Yet the cost of those injuries varies significantly. Manchester City and Liverpool operate with comparable wage commitments, but their estimated injury costs in 2024\/25 were \u20ac41 million and \u20ac15 million respectively (Exhibit 2).<\/p>\n


\nInjury count alone does not explain the difference. Player wage profiles and the type, timing and severity of injuries all matter. But so does the way a department coordinates decisions, progresses rehabilitation phases, manages exposure – and learns from previous cases.<\/p>\n

The arithmetic is simple: Cost = days lost \u00d7 the daily wage of the unavailable player.<\/p>\n

The useful question, then, is not whether every injury can be prevented. It is whether a department can reduce avoidable days lost, limit recurrence and prevent a short absence from becoming a long one. Aston Villa halved its injury bill in a year despite recording three more injuries. Average absence fell from 37 days to 15. This is where the conversation becomes uncomfortable.<\/p>\n

Elite sport has invested heavily in technology, but that alone has not solved the problem. Teamworks, whose platform sits beneath many elite performance departments, estimates that 75\u201390% of collected performance data never influences a decision.<\/p>\n

More feeds. More dashboards. More alerts. But not necessarily more clarity. And certainly not more organisational memory than the room had a decade ago.<\/p>\n

Where AI belongs in the medical \/ performance department<\/strong><\/p>\n

If injury cost is shaped by human decisions, the strategic question is not simply whether AI can predict an injury. It is where, across the work of the department, AI can improve the conditions in which better decisions are made.<\/p>\n

There appear to be approximately seven core stages in a player\u2019s lifecycle, from pre-signing assessment through preparation, availability, rehabilitation, return to play, selection and week-to-week performance (Exhibit 3).<\/p>\n


\nAcross that chain, five technology capabilities matter:<\/p>\n

    \n
  1. Data integration<\/li>\n
  2. Machine learning<\/li>\n
  3. Language models<\/li>\n
  4. AI agents<\/li>\n
  5. Computer vision<\/li>\n<\/ol>\n

    They are not interchangeable. They do different jobs, require different controls and carry different risks (Exhibit 4).<\/p>\n


    \nPlaced against the seven stages, their value is uneven. That unevenness is the point (Exhibit 5).<\/p>\n


    \nResearch into AI\u2019s ‘jagged technological frontier’ shows that the same system can improve performance on one task and degrade it on another that appears remarkably similar. In a large field experiment, consultants working with AI completed suitable tasks faster and to a higher standard. Just outside the technology\u2019s capability boundary, however, they were nine percentage points more likely to reach the wrong answer.<\/p>\n

    In sports medicine, confident error carries a human cost. Summarising a meeting is not interpreting pain. Drafting a case chronology is not diagnosing pathology. Identifying a missing piece of information is not deciding that a player is match ready.<\/p>\n

    The matrix makes this boundary visible. Data integration, language models and agents have potential around return to play, selection, documentation and organisational memory. Machine-learning injury prediction, the most heavily marketed capability, earns a more cautious place. A first-team squad may experience roughly 50 injuries in a season. That is a small, shifting and highly contextual sample from which to make high-confidence predictions.<\/p>\n

    AI is often strong in complicated environments where the rules and relationships can be modelled. Injury is complex. The same load does not create the same response in every player, on every day, in every context, at any point of the season. A model can surface a signal and even prompt a better question. It should not make the call. Judgement should never be passed off.<\/p>\n

    No column in the matrix is empty of people. The technology changes what arrives on the desk \u2013 not who carries the decision. Read together, the lifecycle, the capabilities and the matrix point to three opportunities, in ascending order of ambition.<\/p>\n

    Opportunity one: connect the player\u2019s data to the practitioners<\/strong><\/p>\n

    Here is a precondition for everything else. Most clubs have connected parts of the player record: training load, testing, wellness scores, medical events and match exposure. The practitioner record is less connected. Clinical reasoning, return-to-play criteria, changes of plan, and contextual observations are written up in SOAP notes or similar, filed somewhere and maybe surfaced again.<\/p>\n

    Join both records around one player identity, one timeline and one agreed definition of availability. Then make what has already been written searchable and usable. The challenge is not simply technical. The system needs appropriate consent and permission to read practitioner notes\u2026 not merely permission to store them.<\/p>\n

    Opportunity two: execute the workflows<\/strong><\/p>\n

    Once the record is connected, AI agents can complete or prepare routine work. Our task analysis of the rehabilitation stage within a Premier League backroom department identified 22 recurring tasks, consuming roughly 40 hours each week.<\/p>\n

    Eleven could be produced automatically overnight. Seven could be drafted for a practitioner to review and correct. Two, diagnosis and the return-to-play decisions, should not move from the human. That\u2019s complex terrain.<\/p>\n

    The prize is not a smaller department. It is approximately a day each week returned to the practitioner. More time with players, more time coaching, more time thinking \u2013 and more time discussing the difficult cases.<\/p>\n

    Opportunity three: reinvent how the department learns<\/strong><\/p>\n

    The first two opportunities improve today\u2019s work. The third changes tomorrow\u2019s department.<\/p>\n

    What if the team recorded what it expected at every major decision point? What if it compared expected against actual at the end of each case? What if similar cases, changes of plan and second opinions could be retrieved in seconds? What if the knowledge of an experienced sports med physician, for example, became part of the department\u2019s memory rather than leaving when that person did?<\/p>\n

    This is the move from digitising activity to building organisational intelligence and retaining and managing knowledge. It requires more than a new tool or workflow. It requires the will to redesign the week around better systems, not bolt another platform onto an already crowded one.<\/p>\n

    The importance of strategy<\/strong><\/p>\n

    The question most departments ask is, \u201cWhich AI tools should we buy?\u201d. What about, \u201cHow should our department evolve because AI exists?\u201d That mindset shift matters.<\/p>\n

    Strategy should begin with a performance problem worth solving. Avoidable games lost to injury, inconsistent return-to-play decisions, excessive admin, or lessons that disappear when a case closes.<\/p>\n

    It should not begin with a product demonstration. A performance department is a human environment. A wrong call is felt by a player and carried by a practitioner. Adoption must therefore be intentional, governed and connected to the realities of player care and performance.<\/p>\n

    We recommend five steps, in order, based on recent use case experiences across elite football environments:<\/p>\n

      \n
    1. Audit<\/strong><\/li>\n<\/ol>\n

      Understand how the department works today. Recurring processes, decision points, systems, people, information flows and data gaps.<\/p>\n

        \n
      1. Benchmark<\/strong><\/li>\n<\/ol>\n

        Compare the department stage by stage against the player lifecycle. Compare its outcomes against an appropriate league and reference set, within the league, and even outside the league\/sport.<\/p>\n

          \n
        1. Roadmap<\/strong><\/li>\n<\/ol>\n

          Decide what to improve first, and which capability is required. Consider the pilot case and how the system will expand once the initial workflow proves useful and gains traction.<\/p>\n

            \n
          1. Boundaries and governance<\/strong><\/li>\n<\/ol>\n

            Define what AI may produce, what it may only assist and what must remain human-led. Establish the controls around access, consent, review, retention, accountability and escalation. This step is what separates an AI strategy from a technology project.<\/p>\n

              \n
            1. Monitoring and operating model<\/strong><\/li>\n<\/ol>\n

              Track whether the change improves the work. Decide what measures matter, which capabilities the department needs, and how responsibilities, meetings and incentives must evolve.<\/p>\n

              The jagged frontier must be mapped at task level for each department:<\/p>\n