If artificial intelligence is writing your report, why are you required?
Malone S, Buchheit M. If artificial intelligence is writing your report, why are you required? Defending the CRAFT of performance support in the ever-growing age of automated analysis. Sport Perf Sci Rep. 2026; Sep; #317:v1.
Full Text here
New piece with Shane Malone, and it is the one I have been arguing about in private for a year.
Here is the honest position. Most of what I was taught to be good at is now automatable, and a lot of it is already automated. Cleaning, merging, typical error, smallest worthwhile change, z-scores, the first draft of the narrative. If my value proposition is that I can do those things well, I am competing with something that will do them faster than me next month, and cheaper, and at 3am.
That is not a loss. It is a relocation. The part of the job that could be written down has been taken. What is left is the part that could not: knowing the squad, knowing the coach, knowing which two numbers matter this week, and being willing to put your name on a recommendation and defend it in a room where someone with more authority disagrees.
Shane led this and built the CRAFT structure around it. Five stages, ownership named at each one, and the failure mode when you get the ownership wrong. Delegate where output can be confirmed and error is visible. Keep what needs judgement, context and accountability.
Two things I would underline.
The evidence does not say that adding a human to AI always helps. The meta-analysis we rely on found losses on decision tasks, and the direction depended on whether the human actually knew something the model did not. If you cannot say from memory which player is managing a calf this week, sitting between the model and the coach adds nothing and may subtract.
And the deskilling risk is real, not theoretical. A junior practitioner who has never manually cleaned a GPS file may never build the pattern recognition to spot the velocity spike when a winger gets pushed over the try line. That knowledge is only accumulated by being there.
This extends the POCKET work Shane and I published in SPSR (#229) and connects to the monitoring quadrant with Karim Hader (#258).
One deliberate choice worth mentioning. We wrote an AI Use Statement into the paper, structured by our own model: AI at the Framing stage only, after the conceptual work and the critical appraisal. If we are going to publish a paper telling people where AI belongs in a workflow, we should show ours.
#SportScience #ArtificialIntelligence #PerformanceSupport #KnowledgeTranslation #Leadership
