The most efficient employees in your company are probably already using AI. You don't know about it because they're afraid of what happens if you find out.

An operations leader at a 70-person agency described it to me recently: "Some people, I don't know if they're afraid to admit the efficiencies. They're using it and they're becoming way more efficient. So then they're going on an hour walk with their dog in the afternoon."

Call it the dog-walk economy. The most productive people in the organization are nowhere to be seen at 3pm. Their work is done. They figured out how to compress an 8-hour day into 4. And they tell no one.

Broken organizational incentives produce a specific hidden behavior: capable employees building a closet for their AI-driven efficiency and locking it shut.

Three things employees do with AI gains they can't admit

1. They hide it. A team member at a staffing firm told her director, in an AI training of all places, that "AI can automate everyone else's job around me" — meaning hers is uniquely irreplaceable. The director, sitting next to me, recognized the pattern. Another director on the same team told me his most capable people were silently racing through their workload. In the meetings about AI, the people who didn't want to use it did most of the talking.

2. They pay for it themselves. A commercial lines broker at a 200-person insurance company paid for personal ChatGPT out of pocket because the company restricted AI tools. He built automated renewal notifications and premium calculation workflows on his personal account, outside official IT, for clients of the company that wouldn't let him use AI inside it. The company is paying him to do work that an unofficial system is actually doing — and they have no visibility into the system.

3. They look busy. When the work takes 30 minutes and the calendar block was 4 hours, you don't return the time. You take the dog for a walk. You start the next task slowly. You leave Slack on green. The behavior is protective. Employees preserve what they earned because they have no reason to believe the organization would let them keep it.

Why this is rational

Look at it from the employee's seat. They've watched what happens when efficiency gets visible. The fastest person on the team gets the next hard project, no extra pay. The team that finishes early gets reorged. The person who automated their reporting gets asked to automate someone else's. Sometimes that someone else gets laid off six months later.

A workshop participant from a corporate environment summed up the trust gap in one sentence: "We turned on Copilot and that's what you use. People were secretly using ChatGPT, but there isn't the support for the staff." Top-down tool mandates push people underground. Tell someone they have to use the company-approved tool and they will use whatever actually works — on their own machine, with their own credit card, off the company network.

Language traps leadership before they even start. The same operations leader at the marketing agency told me: "Every time you call it an audit, it freaks people out. They think you're going to change things and let people go." Leaders use audit vocabulary to investigate AI use; employees hear layoff vocabulary. The closet door gets thicker every time someone says the word audit.

So employees do the math. Visible efficiency = more work, more suspicion, or threat. Hidden efficiency = an hour with the dog. The dog wins.

The leadership cost

You cannot capture what nobody admits to. You cannot scale what nobody documents. You cannot study what nobody talks about. The shadow AI economy means your organization is paying for adoption it can't see — and getting almost none of the compounding value that comes from sharing what works.

The broker building renewal automation on personal ChatGPT? That workflow could become a department-wide system if anyone knew it existed. It won't, because telling leadership invites questions he doesn't want to answer. The person walking the dog at 3pm? Their compression technique could save their team 200 hours a quarter. It won't, because the moment they share it, the dog walks become someone else's billable hour.

The closet has a real cost: every individual win stays individual. The organization gets none of the network effects. You're paying full price for AI tools and capturing maybe 10% of the actual productivity already happening inside your walls.

What opens the closet

The fix is an environment where admitting efficiency is safer than hiding it. The work is change management, and it starts the same way every real diagnosis starts: structured listening, not surveys.

Specifically:

Make a public commitment about what saved time is for. Before you ask anyone to share their AI workflows, answer the question they're actually asking: if I tell you I save 10 hours a week, what happens? If the answer is "more work" or "we'll see," nobody will tell you anything. If the answer is documented and credible — new projects, skill development, a clear reinvestment plan — you've cracked the door.

Reward the share, not the secret. Recognize the person who automated their reporting. Make their workflow visible. Give them a title or a project that signals "this is the kind of behavior we promote." If the only thing employees see rewarded is being busy, they'll perform busy and hide everything else.

Stop top-down tool mandates that ignore what people actually use. Half the company on personal ChatGPT is data: the approved tool isn't doing the job, or the support isn't there, or both. Punishing the workaround pushes it deeper into the closet.

The leader's actual question

Most leaders walk into AI adoption asking "how do we get our people to use AI?" Your people are already using AI. Ask instead: "have we made it safe to admit?"

If the answer is no, you're paying for tools, training, and licenses while the actual productivity sits in a closet you built with your own vocabulary, policies, and incentive structures. The dog-walk economy is the receipt.

The leaders who close the gap make it safe to come out.

CitizenWorks helps organizations open the closet — starting with an AI Pulse Check that surfaces the real picture of what's already happening inside your team.