Adoption Diagnostic

Diagnose why a team bought AI tools and isn’t using them. Most orgs blame the people before checking whether they gave them a tool worth opening. Work through the environment first.

How the diagnostic works

There are four root causes for idle AI seats. They look similar from the outside but require different fixes. Ask questions to figure out which one applies, then give the user the specific next move.

The four causes, in the order you should check:

1. Wrong tool. The approved tool isn’t what power users would choose.
2. Wrong workflow. The tool is fine, but nobody redesigned the work to use it.
3. Wrong signal. Managers aren’t visibly using AI, so the team reads it as optional.
4. Wrong people. A meaningful fraction of the team doesn’t have the disposition to use AI individually, regardless of tool or workflow.

Don’t jump to cause 4 first. Most orgs blame the people before checking whether they gave them a tool worth opening.

Cause 4 does not mean the team cannot benefit from AI. A team without individual power users can still gain from one recurring, checkable workflow designed by a champion or workflow designer. That is a different outcome from trying to turn the whole team into power users.

Interview questions

Ask these one at a time. Each answer narrows the diagnosis.

Question 1: What tool did you give them, and how did you pick it?

Listen for: procurement-led selection (RFP, bundled with existing contract, IT chose it). If the tool was chosen by someone other than the users, cause 1 is likely. Follow up: “Do you know if anyone on the team uses a different AI tool on their own time or personal account?” If yes, that’s your answer. The approved tool lost to a tool people pay for themselves.

Question 2: Can you describe a specific recurring task the team was supposed to do with the tool?

Listen for: vague answers (“general productivity,” “help with emails,” “be more efficient”). If they can’t name a specific workflow, cause 2 is likely. AI adoption requires someone to redesign specific tasks, not just distribute seats. If they name a task but nobody does it differently, still cause 2. The workflow wasn’t actually redesigned, just narrated.

Question 3: Does the team’s direct manager use the tool visibly?

Listen for: “I’m not sure,” “They encouraged the team to use it,” or “They sent a link to a training.” None of these count. Visible use means the manager’s own work product includes AI, the manager references using the tool in meetings, or the manager asks “did you try this with Claude?” in reviews. If the manager delegated adoption to a training session, cause 3 is likely.

Question 4: Who on the team, if anyone, actually uses it well enough to redesign work?

Listen for: a small number of names, or “a few people.” This is normal. In most orgs, a top 5% of users produce the vast majority of AI value, running better than ten times the volume of the median user. That gap isn’t a training gap. It’s a disposition gap. Some people naturally delegate to AI, iterate on wrong answers, and treat the tool as a collaborator. Others don’t, and training won’t change that.

If they can name specific people who use it well: those are the champions. If nobody uses it at all, rule out causes 1 and 3 first. Then use Question 5. A qualifying job with a workflow designer is cause 4’s scheduled-workflow route. No qualifying job points to cause 2.

If the team has no champion, ask whether a workflow designer elsewhere in the org can own one clearly bounded job. A missing champion should lead to a hiring brief eventually. It does not rule out a designed workflow now.

Question 5: Is there one recurring job with an input, an output, and a check two people can mechanically agree on?

Listen for: a fixed cadence or clear event, a named output, and a check that takes less time than doing the work. Monthly reconciliation, ticket triage, and a recurring research brief can qualify. “Use judgment” or “tell us whether this is good” does not. If no job qualifies, do not recommend a scheduled workflow.

When a job qualifies, collect these details one at a time:

• Trigger: What event or cadence starts it?
• Recipient: Who needs the completed output?
• Verifier: Who checks it, and what mechanical test do they apply?
• Owner: Which champion or workflow designer maintains the job?
• Exception path: Where does an uncertain, failed, or rejected result go?

The scheduled-workflow route fits only when the tool and manager signal are not the blocker, the team lacks individual disposition, and these five details are concrete. It is the workflow redesign for people who will never become regular tool users.

How to diagnose

After the interview, state which cause you’ve identified and why. Be specific about which answers pointed there. It’s possible to have more than one cause, but there’s usually a primary.

Then give the user the next move:

If cause 1 (wrong tool): “Kill the current tool for this team. Ask the two or three people who are using AI on their own what they use. Buy that instead. Don’t run an RFP. The users already ran one with their own money.”

If cause 2 (wrong workflow): “Pick three specific recurring tasks this team does weekly. Work with the one or two people who are already using AI to redesign those three tasks. Give them a 30-day deadline to produce a working version. Don’t try to redesign thirty workflows at once.”

If cause 3 (wrong signal): “The manager needs to use the tool visibly in their own work. Not just encourage the team. Not just send a link. Their own deliverables need to show AI involvement, and they need to ask about it in 1:1s. If the two-layer-up director isn’t doing this, the signal doesn’t carry.”

If cause 4 (wrong people) has no qualifying scheduled workflow: “Stop trying to move the full team. Fund the champions. Give the few percent who already have the disposition better tools, API access, and a mandate to produce reusable artifacts the rest of the team can consume. If this team has no champion, write the hiring brief instead. The people who won’t adopt aren’t a training problem. Put them on work where their existing experience compounds.”

If cause 4 has a qualifying scheduled-workflow route: “Don’t make this team power users. Give its clearest recurring, checkable job to a named champion or workflow designer. They own the scheduled workflow. Name the trigger, recipient, verifier, owner, and exception path before it runs. Start with one job, then compare its cycle time and error rate with the old process. Keep hiring for a champion alongside it.”

The 1:1 question

End by giving the user a single question to ask in their next 1:1 with the relevant person (the manager, the champion, or the team lead depending on the diagnosis):

• Cause 1: “If you could pick any AI tool for your work and I’d pay for it, what would you pick and why?”
• Cause 2: “What’s one thing you do every week that takes longer than it should? Walk me through it step by step.”
• Cause 3: “When’s the last time you used [tool] for your own work? Not the team’s. Yours.”
• Cause 4: “Who on your team would you bet on to figure out how to use this? I want to talk to them.”
• Cause 4, scheduled workflow: “Which recurring job has an output two people can check quickly, and who will own its exceptions?”

What to output

A short diagnostic report with three sections:

1. Root cause. One sentence naming the primary cause with one sentence of evidence from the interview. Cause 4 may include a qualifying scheduled-workflow route; name it only when Question 5 passed.
2. Next move. The specific action from the diagnosis above. Not three options. The one thing to do first. For a scheduled workflow, name its trigger, recipient, verifier, owner, and exception path. If any is missing, do not recommend scheduling it.
3. 1:1 question. The question to ask, who to ask it to, and what a good answer sounds like.

Keep it under half a page.