There is a lazy assumption moving through football at the moment, and it deserves a proper answer rather than a defensive one. The claim is that anyone using artificial intelligence in their analysis has quietly handed the actual decisions over to a machine. I understand why the idea has taken hold, because the technology is new and the loudest examples tend to be the worst ones. So I want to set out, plainly and without the hype, how I actually use AI across my scouting work. The honest version is far less dramatic than the headlines suggest, and it is a great deal more useful in practice.
AI is the tool I reach for, and I remain the person making every judgement that carries any real weight. Nothing I send to a club is a machine’s opinion that I have quietly dressed up and presented as my own work. Every conclusion in every report is one I am ready to stand behind, defend in a room, and own if it proves wrong. The technology helps me arrive there faster and present it more clearly, but it does not arrive there in my place. Let me show you what that actually looks like across the two places where it earns its keep in my workflow.
Building the pipeline
Good scouting runs on good data, and good data has to be gathered cleanly before anyone can begin to read it well. A while ago I set out to build a data scraper that would pull the raw numbers I needed into a workable form. I am comfortable working in Python, though comfortable is not the same as fast, and my first version behaved as most first versions do. It mostly worked, broke in a few odd places, and occasionally produced a result that looked correct until you checked it properly. This is the point where I used Claude, and it is worth being precise about what that collaboration actually involved in practice.
I did not ask the tool to build me a scraper and then simply accept whatever happened to come back from it. Instead I sat with the code, found where it was failing, and worked through each problem with the AI as a tireless partner. I would describe a bug, test the fix it suggested, reject the ones that did not hold up, and keep moving forward. When the logic was sound but the structure had grown messy, I had it help me tidy the code for later maintenance.
The important thing here is that I was driving the entire way, with the tool reacting to my direction rather than ever setting it. The AI is quick and it never tires, which makes it a useful partner for exactly that kind of slow, iterative work. It does not know which numbers matter for a winger as opposed to a holding midfielder, and it never will on its own. It does not know which quiet edge case will ruin a dataset three weeks later, whereas that happens to be precisely my job. The refining was a conversation throughout, and I was always the one deciding when an answer was finally good enough to ship.
Getting the findings down clearly
Here is something my PhD supervisors would confirm without a moment’s hesitation, which is that I tend to think at considerable speed. Ideas arrive faster than I can put them into any order, and my first draft of anything reads like the inside of my head. That habit matters in scouting more than most people assume, because they imagine the work as little more than watching football. A report is not only a record of what you saw, because it is really a tool for someone else to make a decision. A sporting director or a head coach has no time to untangle an argument that is clever but hopelessly tangled in its delivery.
They need to know what you found, why it matters, and what you would recommend, arranged in an order that respects their time. This is the second place where I bring in AI, and once again it is a matter of command rather than any kind of handover. Once I have done the watching and the thinking and reached my conclusions, I use Claude to help structure and sharpen the delivery. I get my findings down first, then work through the draft to tighten the logic and make sure the argument genuinely lands. The judgement remains entirely mine, while the clarity is something I am content to get help with, because clarity is part of the product. The reports leave my desk under my name, so they have to read as cleanly as the analysis that produced them.
Choosing the right tool
People occasionally ask why I use Claude in particular, and the answer comes down to two things, both rooted firmly in my own standards. The first is the quality of the writing, which is the strongest I have found for the structuring and sharpening work described above. When your output is words on a page that another person has to act on, that quality stops being a luxury and becomes essential. The second reason is the way I think about confidentiality, because scouting touches information that is genuinely not mine to broadcast. Target lists, assessments, and the data a club is paying me to handle all demand real care about where that material ends up.
I have made deliberate choices about the tools I use with sensitive work, rather than reaching for whatever happens to be nearest at the time. That care is part of a practitioner’s responsibility as I understand it, and I treat it seriously rather than as an afterthought. I will add one small detail that I think points to something larger about how I approach all of this technology. Claude is my primary tool, but it is not my only one, and when I need generated imagery I use a different tool entirely. I am not loyal to a brand so much as loyal to the result, which means choosing whichever tool genuinely fits the task in front of me.
Where this leaves us
So if you are a sporting director wondering whether hiring someone who uses AI means hiring someone who has stopped thinking, it does not. If anything it means the opposite, because the technology clears the slow and breakable work out of the path of the real work. That leaves more of my time for the one thing no tool can ever do for me, which is the act of judgement itself. The data scraper still needed someone who knew which numbers actually mattered before any of it could be trusted. The report still needed someone to decide what the player was genuinely worth in the context the club was asking about. The recommendation still needed a person willing to put their own name to it, and on every single occasion that person is me.
