The usual list for keeping good people is pay, title, manager, and a path to the next role. When AI comes up at all, we see it filed under software, next to the laptop and the licenses.
That's a mistake, and our prediction is that 2027 is the year it shows up in who stays and who leaves. This is a prediction, not a fact, and we don't have a study to point to. What we have is a set of signals, some of them survey data, and a pattern we keep seeing in the field.
Some people can't go back.
In the organizations we work with, a small group has gone much further with AI than their colleagues, and started before anyone asked them to. We put the figure at roughly 5 to 10 percent of people in any organization. That is our own estimate from this work, not a survey result. They are not necessarily technical. They are curious, high-agency people who point AI at their own work, find out how much better it can get, and keep going. Colleagues notice, and before long they are the ones getting asked how they did it. We call them AI operators, and many companies call them champions.
Once a person has done real work this way, something changes. They know what is possible, and the old way of working stops feeling like normal and starts feeling like a downgrade. Luke Pierce, who runs an AI automation company, put it bluntly on X in October. If AI went away for good, he wrote, "I'd quit and go learn a trade. No chance I'd go back to operating how we had to before AI." We quoted his post with a prediction: "The companies who offer the most access to tools/tokens will attract the top talent. An employee will leave a previously great role over this."
What the early numbers say.
Surveys can't tell us what people will do, but they show which way people lean. In a June 2026 survey of 1,250 office professionals at large companies in four countries, PagerDuty found that 77 percent believe their company's restrictions or policies on AI use are limiting their professional growth or career mobility, and 75 percent said they would be likely to look for a new job that offered better AI skills development. Saying you would look is not the same as leaving, and PagerDuty sells operations software, but it is still a lot of people.
The market is also bidding for these skills. PwC's 2026 AI Jobs Barometer, which analyzed more than a billion job ads, found that postings for jobs requiring specific AI skills grew 69 percent while the overall job market grew 9 percent, and that the average wage premium for AI skills reached 62 percent, up from 57 percent a year earlier. That is an average across the job ads PwC analyzed, not what any one person earns. It does mean a champion who looks around will find employers who already value what they can do.
Two things push champions out.
The first is the lack of a path. A person who builds something useful for their team and gets a thank-you and the same job description is doing a new kind of work in an old role. Nothing tells them it leads anywhere. If nobody asks them to do more of this work, or says what it could turn into, they will eventually find someone who will.
The second is a cap on how much AI they can use. Many AI products meter usage, and the employer controls part of it. On a Claude Team plan, for example, limits apply to each person each week, and the seat type determines the allowance. An organization can enable usage credits so people keep working after they reach their included limit, and it can set spending caps. For an occasional user, a limit may never show up. For someone who has built real work on top of these AI products, it is a ceiling on the job, and they know another employer may not impose one.
Put those two together and you have a person with hard-to-find skills, a good sense of their market value, and a reason to look around.
Why we expect it in 2027.
Today the champions are still early. McKinsey's 2025 survey found that nearly two-thirds of respondents said their organizations had not yet begun scaling AI across the enterprise, so for most companies the gap between people who have gone deep and everyone else is only starting to show. As rollouts reach everyone, the gap becomes visible, and so does the number of employers who understand what these people are worth.
We see some companies building around AI. They offer wide access, real scope, and leaders who use AI themselves. When a champion at a slower company compares that to their current job, the choice gets easy. Our prediction is a large reshuffling in 2027: many of the best AI employees at existing companies moving to companies that embrace AI, with the companies they leave feeling it.
What we don't know.
We can't say how big the move will be or which industries feel it first. Some champions will stay, especially where leaders give them room. People who say they would look for a new job often don't, and the timing could slip into 2028. We're more confident in the direction than in the date.
What you can do now.
You don't need to wait for 2027 to act. These steps are cheap, and they are good management whether or not the prediction comes true.
Find your champions. Ask managers who people go to when they need help with AI. You will likely hear the same few names.
Give them a real path. Make the work part of their job. Name the role, set aside time for it, and tell them what it could lead to. Put them in front of senior leaders, so they can ask for what they need directly.
Check your limits. Look at who is hitting usage caps and why. If the person at the cap is producing real results, raising it is probably the cheapest retention move you have.
Ask what is in the way. A short conversation with each champion about what blocks them will tell you more than a survey. Then ask what a person with their skills could get elsewhere, and decide whether your offer can compete.
For more on finding and supporting these people, see Building an AI Champions Program That Actually Works.
Sources
- Luke Pierce on X (October 2026)
- Brandon Gadoci on X (October 2026)
- PagerDuty, Shadow AI Survey (June 11, 2026; 1,250 office professionals, Wakefield Research)
- PwC, 2026 Global AI Jobs Barometer (June 15, 2026)
- Anthropic, What is the Team plan? (Claude Help Center: per-member weekly limits, seat types, usage credits, spend controls)
- McKinsey, The state of AI in 2025: Agents, innovation, and transformation (November 5, 2025)
- Gadoci Consulting, Services (AI Operators: our own estimate that 5 to 10 percent of people in every organization already use AI)