Ask yourself something before you read another word: do you know which AI tools your team used this week? Not the ones you rolled out and announced in a meeting. The ones somebody found on their own, on an ordinary Tuesday afternoon, because the tool you gave them was too slow, too limited, or didn’t exist yet. If you can’t answer that with any confidence, you’re not facing a technology problem. You’re facing the same problem I’ve spent 35 years and more than 500 organizations working on: culture defaults to the personality of whoever’s running the business, whether anyone designed it that way or not. Right now, for a lot of owners, that default is silence on AI. And silence gets filled by somebody.
Researchers have a name for what fills it: “shadow AI,” the tools employees use on their own because the approved option is missing, clunky, or doesn’t exist. SHRM (opens in new tab) reports that Gartner projects 75% of employees will use some form of shadow IT by 2027, up from 41% in 2022. That isn’t a fringe habit creeping in at the edges of a few departments. That’s most of your workforce.
Here’s where I’ll be honest with you about where most owners stand, because I hear it from clients every week. About a third of you are fascinated by what AI can already do, though you’ve only scratched the surface of it. Another third are testing the waters, trying a tool here and there, without ever learning to leverage it for real work. And a third of you are frightened of it outright, mainly because you haven’t taken the time to sit down and learn it. Wherever you land in that mix, here’s the part that should get your attention: your team isn’t waiting on you to catch up. They’ve already started.
The employees are the internal customer… automation is there to help us develop relationships… it never can replace, at this point, the human element.
— Gary Henson, Founder of Gary Henson Group
Some owners hear all of this and reach for the obvious fix: write a policy banning unapproved AI tools and call the problem handled. I’ll tell you what I tell every client who tries to solve a people problem with a rule instead of a relationship. It won’t hold. Somebody trying to do their job better doesn’t stop wanting to do it better because you told them not to use the tool that helps. They just stop telling you about it. You’ve traded a behavior you could manage for one you can’t see, and you’ve taught your team that being resourceful around here gets punished instead of guided.
What works here is the same thing that works in every culture rebuild I’ve ever run. You don’t fix a vacuum by policing it. You fix it by filling it yourself, on purpose, before somebody else fills it for you. Sit down and decide what you genuinely want your people using AI for, then put it in writing the same way you’d write a job description or a company commitment: one page that functions as your AI policy, covering what’s approved, what information never leaves the building, and who they ask when something new comes up. Microsoft and LinkedIn’s 2024 Work Trend Index (opens in new tab) found that 78% of AI users are already bringing their own AI tools into work rather than waiting on a company to catch up. Your people aren’t the exception to that number. They’re the rule.
of AI users say they already bring their own, unapproved AI tools to work rather than waiting on their company to catch up. Microsoft and LinkedIn's 2024 Work Trend Index →
Once that page is written, the next move is a conversation, not a company-wide announcement: one on one, with the people who report to you, asking what they’re already using and why. Treat the answers as information for building ground rules that fit how your business runs, rather than as evidence for a write-up, and you’ll probably find people using AI for things you’d have approved immediately if they’d asked, alongside a few things worth redirecting. Either way, you can’t set expectations for behavior you don’t know is happening.
This is the same test I put every client through before we ever touch an org chart. Your employees are the internal customer of this business, and that relationship runs both directions: you expect their best work, and they’re entitled to expect that you’ve told them plainly what “best” looks like. Bring shadow AI into the light and it becomes exactly that: a tool your team already wanted to use well, that you finally gave them the structure to use well. Leave it in the shadows and you’ve added one more thing you don’t know about your own business, at a moment when you can least afford another one.
GH Group has covered the other half of this same problem, checking what AI hands back to you before you act on it, in The Tasks Small Business Owners Shouldn’t Hand to AI Yet, and the same discipline of putting expectations in writing shows up again in Writing Company Commitments Employees Will Remember. The Inside-Out Method™ starts in the same place either way: not with the technology, but with the owner deciding, on purpose, what they want.
