Approved, But Nobody's Using It
The last piece covered "IT deploys it, but who runs it?" — the question of AI ownership. Say you've designated an owner and cleaned up the org chart. Then a familiar scene follows. Two months after signing the deployment sign-off, you open the dashboard: the only active user is the executive who approved the project.
The tempting conclusion is "employees hate change." I used to think so too. You add more training, resend the internal memo, tack on "mandatory use from now on" — and adoption still won't budge. This is where the framing has to shift. The problem isn't the people. It's the workflow path.
The Cost of Misdiagnosing It as "Resistance"
Diagnose it as change resistance and the prescription writes itself: more training. But no matter how much training you add, it barely correlates with adoption. A recent Korean survey found 69.5% of companies had adopted AI, but only 12.5% had it embedded as a work standard. That's a 5.5x gap between deployment and adoption. If the gap were purely down to "insufficient training," no company on earth would still be wrestling with adoption.
US surveys point the same way. 85% of companies increased AI spending in the past year, but only 6% saw meaningful results in that time. And just 27% actually integrated the tools into their daily workflow. The other 73% left the tool "next to" the work. A tool sitting next to the work doesn't get used — the same way the umbrella by the door gets left behind when you walk out.
The Real Cause — The Moment You Open a Second Window, You've Lost
The pattern I see on the ground usually looks like this.
A sales rep lives inside the CRM all day. Next to it, they're supposed to open a separate browser tab for the AI tool. Get a summary, copy it, paste it back into the CRM. That "one extra tab" is what resistance actually is. Time it and it's 15 seconds per pass, two and a half minutes across ten uses a day. Not a huge number, but daily friction quietly eats adoption alive. And people can't articulate why they don't use the tool. Their hand just doesn't reach for it.
Manufacturing back-office work is similar. 90% of PO verification happens inside Excel. Then a review AI ships that requires logging into a web dashboard. "You only need to log in twice a day" doesn't work. It's not that people refuse to log in twice — it's that logging in and porting results back into Excel is a motion that wasn't there before, so the muscle never forms.
What It Means to Move Inside the Workflow
There's only one direction that works. You don't bend the person around the tool. You push the tool into the person's day. In practice this splits into three forms.
First, layer it onto the screen they already use. An AI summary card just sits at the bottom of the CRM detail page. The user reads it without ever thinking "I'm using AI." Adoption metrics become meaningless because usage is just the default state.
Second, embed it inside the messenger they already use. Access via a bot on Slack, Teams, or Kakao Work. No separate app means no login, no context switch, no window juggling. One mid-sized company opened a single "ask the bot about expense reports" channel and hit 40% usage in the first month — eight times higher than the version that lived in a standalone portal.
Third, run it inside the Excel and documents they already have open. This is exactly why the Copilot family is pushing in this direction. It's not perfect, but the mere fact that results appear in "the file already open" drops early resistance substantially.
Three Questions to Ask Before You Deploy
Honestly, these three questions belong on the desk before deployment is approved. Asking them later isn't fatal, but six months after sign-off the redesign cost overtakes the original design cost.
First: how many times a day does the target user do the task this tool is replacing, and inside which screen? Second: once the tool produces output, which system does that output need to flow into next? Third: how many clicks does that flow take? If the answer to the third is "three or more," the tool typically dies on the shelf. Three clicks doesn't beat a habit.
What's Coming Next
Say you've succeeded in putting the tool inside the workflow. Now the person who used to do that task has time back. What fills that time is the next problem. Reframing it from "people become surplus" to "the work itself changes" — the next installment covers the practical side of role redesign.
If you want the earlier piece in this series, "IT Deploys It, But Who Runs It? — The Position of the AI Owner" reads better first. The "where do we put it" conversation in this piece only really begins once "who runs it" is settled.
If "we deployed it and no one uses it" keeps coming up internally, feel free to book a workflow-redesign consultation. This isn't a conversation about swapping tools — it's about redrawing the spot where the tool belongs.