If your organization hasn’t decided how employees should use AI, there’s an excellent chance your employees have handled that problem for you.
Jane has ChatGPT turn meeting transcripts into follow-up plans. Luis uses Claude to review proposals before they go to clients. Priya has built something elaborate involving a Project, two AI agents, and a prompt she refuses to share because it took her three months to get right. Meanwhile, their managers are thinking “using AI” means using ChatGPT instead of Google.
Congratulations. You have an AI operating model. You didn’t choose it, and nobody’s really managing it.
The unofficial system usually arrives first
This is a pretty normal way for workplace technology to spread. Someone discovers that AI can save twenty minutes on a task they hate and tells a colleague. Before long, there are six different versions of the same trick floating around the department. Nobody called a kickoff meeting or wrote a rollout plan. People were just trying to get their work done.
So debates about whether employees should use AI can be oddly disconnected from what’s already happening. In many workplaces, employees have decided for themselves. They use it when it helps, avoid it when it doesn’t, and often don’t talk much about either choice.
Along the way, habits take hold. A few are probably terrific. Others, less so. Most managers have no idea which is which.
A policy only gets you so far
Most organizations eventually create rules for AI, which makes sense. People need to know what information is confidential and which systems they’re allowed to use. But those rules don’t tell people how to use AI well.
A policy can say, “Don’t paste sensitive information into an unapproved tool.” It probably won’t tell someone how AI fits into preparing a memo to the CEO or whether it should be part of planning a client meeting. Employees are usually left to work that part out on their own.
The company has said, “Here are the boundaries.” The employee’s still wondering, “Fine. So what am I actually supposed to do with this thing?”
Ask how the work is actually getting done
Imagine their manager asks a simple question at the next staff meeting: “How are you using AI to prepare for client meetings?”
Jane explains what she’s been doing and, actually, it’s pretty good. Luis has a different approach that catches a few things Jane misses. Then Priya opens her three-month-old creation and everyone realizes nobody’s reproducing that thing without an instruction manual and possibly snacks.
Suddenly they’re not discussing AI in the abstract anymore. They’re looking at work they already understand. Jane’s method might be a perfectly good starting point for the team. Priya’s contraption may have one brilliant idea buried inside it. And somewhere in the conversation, the manager can settle which tools are appropriate for this work and what needs to happen before anything reaches a client.
That beats announcing that the company is “embracing AI,” which sounds impressive right up until someone asks what everyone’s supposed to do differently on Tuesday.
Don’t standardize the life out of it
None of this means every employee needs the same prompt or an eight-step AI procedure. Please don’t do that. We’ve already got enough SOPs that stop being followed after two weeks.
But you also don’t want everyone inventing their own system forever. Your employees have already spent months experimenting. Find out what they’ve learned before deciding how the team should work going forward. The useful ideas can spread, the bad ones can die, and there’s still room for somebody to come up with something better next month.
Whatever you decide has to survive Tuesday afternoon
If using an approved AI tool to draft an agenda requires a process longer than drafting the agenda yourself, employees will make a fairly predictable decision about the value of that process.
Managers need to understand enough about how AI is affecting the work to manage it, while employees need room to keep finding better ways to use it without feeling as though admitting they used AI is some sort of confession. A policy won’t get you there by itself. You have to look at the work people actually do and decide what good use looks like there.
Your employees are already building the system. The question is whether it stays accidental.