Issue No.70: The Practitioner Paradox
The Signal
AI researchers are observing that the better AI agents become at generating output, the more exhausted humans feel trying to manage them. This is specific to those working as developers and engineers but I believe this also resonates with those working in human performance.
Physiotherapists, strength coaches, and performance analysts face dashboards displaying heart rate variability, GPS load, sleep quality, menstrual cycle phase, strain scores, recovery metrics, and training readiness algorithms - to name a few. The data arrives in real-time. Decisions about player selection, training modifications, and injury risk must happen immediately.
The tools promise optimisation. The reality is decision fatigue, especially if the technology generates decisions faster than human cognition can process them.
The Strategic Lens
The shift from data collection to decision-making represents a significant change in sports performance technology. The question is no longer “can we measure this?” but “who decides what the measurement means?” This evolution has happened in three phases:
Phase 1: Data scarcity (2000-2015) Practitioners had too little information. GPS units were expensive. Heart rate monitors were basic. Sleep tracking didn’t exist. Injury risk models relied on training logs filled out by hand. Decisions were made from experience and intuition because data simply wasn’t available.
Phase 2: Data abundance (2015-2022) Wearables proliferated. Every player wore multiple sensors. Cloud platforms aggregated streams. Dashboards displayed everything simultaneously. The problem inverted and we had too much data but not enough interpretation, risking practitioners spending more time looking at screens than athletes.
Phase 3: Decision interfaces (2022-present) AI algorithms can now interpret data streams and output recommendations. For example, “Player X shows elevated injury risk - reduce training load by 20%.” “Athlete Y’s recovery score suggests readiness for high-intensity session.” “Menstrual cycle phase indicates strength training optimisation window.”
This sounds like progress as the technology now does the analysis and practitioners make the final call. Human judgment augmented by machine intelligence.
But here’s a paradox that is worth exploring: the more sophisticated the decision support becomes, the more cognitively demanding the practitioner’s job gets.
Why? Because each AI recommendation requires evaluation. Is this algorithm validated for our population? Does the model account for contextual factors? What happens if I override the system and I’m wrong? Should I trust the readiness score or the athlete’s self-report? Which data stream takes priority when they contradict?
Decision-making has always been hard, and more data doesn’t automatically make it easier. Instead, it risks making practitioners responsible for validating and contextualising algorithms - all while maintaining the same decision velocity they had with less information. This highlights a bottleneck effect of human cognitive bandwidth.
The Practitioner Perspective
From my experience as a Physiotherapist, data helped demonstrate objectivity and improve engagement but on a personal level, I have experienced times when recovery metrics from wearables suggested a low score but I felt great.
Who do you believe - the algorithm or how you actually feel?
The “right” answer depends on context that the data can’t capture. Did you sleep poorly because of stress about a family issue, or because of physiological overload? Is the elevated heart rate variability a sign of parasympathetic fatigue, or did you have coffee an hour ago?
These aren’t questions tech answers. They require judgment, pattern recognition, and intuition developed through clinical experience. But the presence of algorithmic recommendations risks creating pressure to justify every deviation.
This is where decision fatigue compounds - when you are having to make contextualised judgments about algorithmic outputs.
The Commercial Reality
The sports performance tech market positioned itself on a promise: more data leads to better decisions, leads to competitive advantage.
This wasn’t wrong to believe. The measurement technology is remarkable, and improving all the time when it comes to validity and reliability.
But the value proposition assumed that practitioners had unlimited cognitive bandwidth to process information streams. It assumed that decision quality scaled linearly with data quantity. It assumed that the constraint was lack of information, not capacity to act on information.
Those assumptions are now becoming questionable.
What we’re seeing now is companies that sold “data platforms” are pivoting to “decision platforms.” The pitch shifts from “track everything” to “tell me what to do.”
This sounds like progress - take the interpretation burden off practitioners and let AI handle it. But this creates new commercial tensions for practitioners where technology threatens to replace human expertise, for organisations if practitioners don’t trust the tech to act on recommendations, and for tech companies if product roadmaps faces contradictory demands.
Female athlete tech faces an acute version of this problem due to the systematic research gap and algorithms therefore being validated on male populations.
This poses capital deployment questions: invest in research to validate algorithms properly (years, expensive, uncertain outcome), or ship products with wellness positioning and population-level correlations (fast, cheaper, commercially viable but scientifically incomplete)?
The Path Forward
The sports performance tech bottleneck becomes a human performance problem that requires an understanding of how practitioners actually make decisions under pressure.
This doesn’t mean more data or more sophisticated algorithms. It might mean though systems designed for the speed humans can actually process and contextualise recommendations, algorithms that communicate uncertainty explicitly, systems that learn from practitioner overrides rather than treating them as errors, and allowing AI to handle data processing and pattern recognition but ensure humans make contextualised judgments about what actions to take.
In essence, this means companies building decision support systems to solve for practitioner cognitive capacity rather than technological capability.
The Takeaway
Sports performance tech has followed the same trajectory as AI agent adoption: initial productivity gains, then cognitive overwhelm, then humans becoming bottlenecks in systems designed to augment them.
There is a need for “mental fitness of an athlete” to manage fleets of AI agents and the outputs. Because more data doesn’t equal better decisions. It equals more decisions, and decision quality depends on cognitive capacity to process, contextualise, and apply information.
Ultimately, the question is whether tech can be designed to work at human speed because that’s the only speed at which good decisions actually get made.
Sometimes the most advanced technology is the one that lets humans be human.
I would love to know your thoughts on this Issue by commenting below.
Thanks for reading! Nic x