What Generative AI Will Never Know About Your Locker Room

What the model never feels in the locker room

Generative AI is here to stay in people functions: it drafts performance reviews, summarises climate surveys, suggests feedback messages, even helps write locker room speeches. It saves writing time. What it cannot do is know the real story of that team.

• Generative AI writes well, but it was never in the room when things actually happened.
• A generated text can sound perfect and say exactly what nobody needed to hear.
• The risk isn't that AI writes worse than a person, it's that it writes without knowing the context.

Human context no prompt can capture

We've seen performance reviews generated almost entirely by AI, flawless in language and completely off the mark about what actually happened with that person that year. We've also seen internal announcements written by a language model that read well on screen and landed badly in the room, because they ignored a recent tension in the team or an uncomfortable decision that still stung.

The problem isn't the tool, it's delegating a task that requires accumulated human context: memory of past conversations, the right tone for that exact moment, sensitivity to what isn't said out loud.

• Performance reviews with flawless wording and generic observations that could apply to anyone.
• Internal messages that ignore a recent tension because the model simply didn't have that information.
• AI-drafted feedback that sounds good but that the manager themselves couldn't defend in a real conversation.

AI summarises; the locker room decides

In sport, a good locker room speech never comes from a template. It comes from knowing each player, what's happened in recent weeks, what the squad needs to hear at that exact moment. Generative AI can help organise ideas or polish a draft, but the judgement over what to say and how to say it still belongs to whoever actually knows the people in the room.

That's why we defend a very specific use of these tools: as a working draft, never as the final product; as support for thinking faster, never as a substitute for whoever holds the real context of the people the message is aimed at.

Where to use it and where not to hand over judgment

Bringing generative AI into internal communication and evaluation processes can work well, as long as clear rules exist from day one.

1. Use it for the first draft, never for the final version
   • Let it draft a structure or a first pass, but always have someone with context review it before it goes out.
   • The more sensitive the message, the more weight human review should carry.
2. Feed it real context, not just generic data
   • A climate summary generated without knowledge of recent events can sound flat or simply wrong.
   • Always include context notes that the model has no way of inferring on its own.
3. Keep the most delicate messages for human writing
   • Announcements about restructuring, underperformance or a change in direction should always originate from a person.
   • AI can help polish the tone afterwards, not decide the message itself.

Questions before asking a chat for advice

Before letting generative AI write on behalf of your organisation, it's worth asking:

• Who reviews AI-generated texts before they reach the team?
• Does the model you use actually know the recent context of the people the message is aimed at?
• Could you defend every sentence of an automatically generated report in a face-to-face conversation?
• Are you using AI to think faster, or to avoid thinking altogether?
• Which messages in your organisation should never come from a machine, no matter what?

Let's map what must stay a human conversation

We're not against generative AI, we use it too. We're in favour of using it while knowing exactly what it cannot do for you.

If you'd like to look at how generative AI is entering communication in your organisation or club, let's talk. It's a simpler conversation than it seems.