Will AI Replace Sports Coaches?

Five forty in the evening, standing next to the car, twenty minutes until practice. The phone has a notification saying Saturday’s footage has finished processing and two hundred and twelve moments have been tagged and sorted by player.

Two hundred and twelve. You have twenty minutes and a session to run.

There is a particular kind of unease in that, and it has nothing to do with robots taking the whistle. The unease is quieter, and it goes something like this. If the machine can watch the game better than you can, what exactly are you for?

The job title gives the answer away

The federal government does not have an occupation called sports coach. It has one called coaches and scouts, and those two words have had very different decades.

The Bureau of Labor Statistics counts about 306,500 people in that occupation and projects it to grow roughly 6% through 2034, faster than the average across all jobs, with something like 41,800 openings a year. That is not a profession under threat.

But look at what got automated inside it. Watching hours of footage to find the moments that matter, logging events, building a report on an opponent, counting things. That was scouting, and platforms like Hudl and Veo now do a version of it automatically for a club subscription that a junior team can afford.

The scouting half went first because it was information work. The coaching half grew because it happens in front of people, outdoors, with children who had a bad day at school.

So the growth number and the automation are both true at the same time, which is why this one gets argued about so much.

What AI cannot do on a Tuesday night

Strip the week back and three things stay stubbornly human.

The first is being the adult in the room. Fourteen kids on a wet field in February, two of them not speaking to each other, one who has been crying in the car and is not going to say so. Holding that together for ninety minutes is not a task a model performs badly. It is not a task a model performs.

The second is selection. Any tool can rank the numbers, and the numbers will tell you the more talented player should start. They will not tell you what it does to a squad when the kid who trained every night for six weeks sits again, and they cannot sit across a table on Thursday and defend the decision to a parent who is upset.

And the third is knowing when to say nothing. Whether a player needs a coaching point right now or needs to be left alone for ten minutes is a read built from months of watching one specific person, and getting it wrong is how you lose them for a season.

Something we keep running into at the Workplace AI Institute is that coaches almost never arrive worried about the whiteboard. They arrive worried about the inbox.

The clips are not the problem, the twenty minutes are

Here is where it actually bites, and it is not glamorous.

Automated analysis made the raw material cheap and abundant. It did nothing about the bottleneck, which was never getting the information. It was having half an hour on a Wednesday to turn the information into two sentences a fifteen-year-old can hold while nervous.

That gap got wider, not narrower. Every opponent now has the same tools you do, so having the data stopped being an edge roughly the moment everyone had it. What is an edge is the coach who can walk in on Thursday with three things instead of thirty.

Which is a writing and thinking problem, and one you can hand over in part. Something like this, pasted with your own scrawled notes underneath, takes about ninety seconds.

Below are my rough notes from watching an opposition team, typed exactly as I scribbled them, using jersey numbers rather than names. Turn them into a summary with their two biggest threats, their two clearest weaknesses, and what they do at set pieces. Then give me the three highest-value things my team should do differently, ranked. Anywhere my notes are thin, say so rather than filling the gap.

Pay attention to the closing line about thin notes. Leave it off and you get a confident, complete-looking report, then no way of telling in three weeks which parts you saw with your own eyes and which the tool filled in for you.

Coaches tend to underestimate how much of a week has quietly become admin with a whistle attached. The 3-minute readiness check walks it session by session and message by message, and flags what a tool could already be handling.

The thing this frees up is the thing coaches never get

Now the part that makes this worth caring about.

The Aspen Institute’s National Coach Survey of more than 10,000 coaches found fewer than three in ten had received any training in the previous year, and it was not for lack of wanting it. Project Play’s State of Play 2025 found recent safety and injury-prevention training among youth coaches had fallen to 26% from 34% in 2019.

Coach development is not losing to disinterest. It is losing to registration forms.

So the hours are the whole argument. An evening you do not spend rewriting the same parent email is an evening that can go into a qualification, or an extra session, or your own family, which is the reason a lot of coaches quit in the first place.

If you want the structured version of that, the AI for Sports Coaches course works through session and season planning, scouting summaries, the athlete and parent conversations that are hard to get right, and the fundraising and admin that keep a program alive.

So will AI replace sports coaches?

No. It already took the half of the occupation that was information work, and it left the half that requires standing on a field.

What it changes is what a coach is for. Less of the counting, more of the deciding. Less of the watching, more of the choosing what to say about it. That is a better job than the one before it, and it is only worse if the twenty minutes in the parking lot never get any longer.

Not sure whether your Tuesdays are coaching or admin with a whistle attached? The readiness check will tell you in about three minutes.

Two hundred and twelve clips is not the threat. Nobody on that field needs two hundred and twelve of anything. They need three, and they need you to have picked which three.