3.9/5
Average excitement about using AI in their role.
What 428 anonymous employees told us about why AI isn't sticking inside mid-market companies.
Every number in this report comes from one place: anonymous surveys we run inside companies before we train a single person.
When a company brings us in to roll out AI, the first thing we do is survey the whole workforce. The surveys are anonymous by design. We collect department and general role, nothing personal. No answer can be traced back to a person, and everyone taking it knows that.
That design choice matters more than anything else about this data. Employees do not tell their boss the truth about AI. They worry it sounds like admitting they cut corners, or like volunteering to be automated. Ask in a meeting and you get silence. Promise anonymity and they tell you exactly where they're stuck, what they don't understand, and what they wish someone would show them.
This report is what they said.
428 employees at five mid-market companies: three commercial construction firms, a national consumer-products brand, and a retail-services group. Surveys ran from February through July 2026.
Method notes: five survey instruments share a core question spine, and a few questions ran in three of the five companies. Multi-select questions can add to more than 100 percent. Percentages are of respondents who answered each question.
Most AI announcements are written to calm a fear. Leadership assumes the workforce is scared, so the memo leads with reassurance. The data says the fear barely exists. What exists instead is a guidance vacuum, and it's enormous.
Put those together and the picture is not a workforce resisting AI. It's a workforce standing at the door, waiting for somebody to open it.
Excitement is not your constraint. Asked how excited they are to use AI more in their role, two thirds of all 428 respondents answered 4 or 5 out of 5. The average across every company was 3.9.
Notice the split at the bottom. Usage is bimodal: a third of the workforce runs AI every day while a quarter has never really started. The middle is stuck at "tried it a few times." Which group an employee lands in has less to do with age or role than with one question: did anyone ever sit with them and show them?
Every survey ends by asking whether the respondent wants to help colleagues adopt AI as an internal champion. Nobody gets paid extra to say yes. Yet:
Two out of three employees lean in when asked to help. The raw material for adoption is already on your payroll. In most companies, nobody has ever collected it.
We ask every employee what their biggest challenges are when using or thinking about AI. Here is the full ranking:
Read the top of that list again. Which tool. What to use it for. Can I trust it. How do I ask it. Every one of them is a question that a knowledgeable person could answer in five minutes. None of them is a technology problem, a budget problem, or a willingness problem.
The same pattern shows up when people describe their overall approach. Only 32% actively hunt for ways to use AI on their own. Everyone else is some version of waiting:
And the wish is explicit: 65% say they'd like to learn how to automate part of their work but never have. Another 5% tried and couldn't get it working. The demand is sitting there, unanswered.
Here are two numbers from the same dataset. 37% of employees use AI daily or more. 66% can't say what data is safe to put into it.
Those two groups overlap. A meaningful share of your workforce is pasting work material into AI tools while guessing at the rules, because nobody ever wrote the rules down where they could find them.
At one firm, only 14% knew specific guidelines existed at all, and 68% said their AI use had been entirely on their own. No formal encouragement, no formal rules. Just improvisation.
This is the finding that should worry you more than the job-fear headline ever did. The risk is not that your people will start using AI. They already started. The risk is that they're doing it unsupervised, on personal accounts, with client data, and every one of them is making up their own policy as they go.
The standard corporate answer to the gap is a training event: book the all-hands, run the demo, hand out the cheat sheet. We deliver training for a living, so take this from us. The event alone doesn't hold.
Ask employees how they actually learn new tools and the answer is hands-on, next to someone who knows:
Here's what the preference data can't show you, but our delivery calendar can. After a group training event, usage spikes for about two weeks. Then the questions start. "How do I make it do this with our template?" "Why did it get this wrong?" "Is this allowed?" Those questions arrive on day three and day thirty, not during the session. If there's no one to catch them, each unanswered question quietly converts one employee back to the old way of working.
Across all 428 employees, two averages tell the whole story.
That distance between excitement and confidence is the Adoption Gap. It's the space where AI initiatives go to stall. Every company in this dataset had already bought tools when we surveyed them. The gap survived the purchase, because tools were never what was missing.
What's missing is a person. Someone who knows the tools, knows the work, and is reachable at the moment an employee gets stuck. The data points at this from every direction: the number one challenge is "what do I use it for in my role," 61% wait to be shown, 65% want to build automations they've never built, and the preferred way to learn is hands-on with an expert. Different questions, same answer.
Walk the floor and ask five people one question: "When you get stuck using AI on real work, who do you ask?"
If the answer is a name, you're ahead of nearly every company we survey. If the answer is a shrug, Google, or "I just gave up," you've found your gap. It's not a tools problem. It was never a tools problem.
Each move below maps to a number in this report. Do all five and you're ahead of every company in the dataset.
One page. What's approved, what never goes into an AI tool, and who to ask about edge cases. 66% of employees are guessing at this today, and most are using AI while they guess.
"Not sure which tool to use for what" is the single most-cited challenge (37%). Pick your approved tools, say what each is for, and kill the ambiguity in one memo.
61% of your people are waiting to be shown. Adoption dies in the unanswered question, so make sure every stuck moment has a destination: a named expert, internal or external, who actually answers.
27% of employees raise their hand to champion AI when asked, and another 39% say maybe. Find yours with an anonymous survey, then give them coaching and a mandate instead of leaving them to freelance.
Seats purchased is a spending metric. Track questions asked, questions answered, and hours saved per team. The 37%-use-it-daily number should climb every month; if it doesn't, the gap is winning.
About the data: 428 anonymous survey responses collected February through July 2026 inside five mid-market companies (commercial construction, consumer products, retail services) as part of Northwest AI consulting engagements. Results aggregated and anonymized; no individual company or person is identifiable.
That question is the whole gap. We built our embedded expert service to answer it: our AI experts live in your Slack or Teams, anyone on your team can ask anything, and we build the automations with your people until they can do it without us. Built for companies doing $25M+ a year. And if you want your own company's version of these numbers, the anonymous survey is where every engagement begins.
You don't have to fix all five numbers this quarter. Book one call and we'll tell you which one is costing you the most.
Backed by our 60-day money-back guarantee.