Everyone on your team uses AI. Write the one page that says how.

Two surveys published this month describe the same shape from two very different corners of working life. People use AI a lot. Far fewer have agreed how.

What the surveys found

Help Net Security reported on Gartner’s research into internal audit on September 15. According to that report, 93% of audit leaders and auditors say they use AI at some level. Only 38% of the 161 chief audit executives Gartner polled in May have an AI strategy at any level, and another 39% are building one. Of 142 chief audit executives polled in May, 54% had not started measuring the value of their AI use.

The most useful number is the obstacle. Audit leaders and auditors named unclear expectations for AI tool use as the most common one, at 48%, followed by differences in how individual auditors use the tools. The fix Gartner puts forward is structured use cases for high-value workflows, and 12% of respondents are piloting them.

On September 23, Inside Higher Ed published its survey of 376 provosts. Roughly seven in 10 use AI in their own work at least once a week. Only one in 10 said their institution has a centralized AI strategy. The most common uses were summarizing documents and reports, presentations and meetings, and drafting communications. Only 12% said institutional AI investment had led to department workflow optimization.

Audit chiefs and provosts are not your team. But the pattern is worth noticing: heavy personal use, no shared rules, and little change in how the department itself runs.

Why this lands on the director

A company strategy, when one exists, is written for everyone and fits no one exactly. The actual expectations get set one level down, by whoever decides how the team’s week works. That is usually you.

When nobody sets them, each person invents their own. One analyst pastes client data into a chat tool, another refuses to touch it, and a third sends a report that reads like it was never checked. None of them is wrong by any written rule, because there is no written rule. That is what “unclear expectations” looks like from the inside.

You do not need a strategy deck to fix this. You need one page.

The one-page department playbook

Keep it to a page so people actually read it. Five short sections:

1. Our two workflows. Name the two recurring pieces of work where the team will use AI on purpose this quarter, for example the weekly status report and first drafts of customer replies. Everything else is personal experimentation, which is fine, but it is not the department’s practice yet.
2. What goes in. What material may be used with which approved tools, and what never leaves your systems. Point to your company’s data and security policy rather than restating it. If there is no policy, ask whoever owns security before you write this line.
3. Who checks what. For each workflow, the person who reviews the output before it leaves the team, and what they check for: facts, numbers, tone, anything promised to a customer.
4. How we say it. When the team discloses AI help, to whom, and in what words. Keep it consistent across the department.
5. How we measure it. The time each workflow takes today, measured before anything changes, and the date you will measure again.

A decision rule for adding a third workflow

Add a workflow to the page only when all three are true for the ones already on it:

• The reviewer can tell a good output from a bad one in a few minutes.
• The time measured after the change is lower than the baseline, and quality complaints have not gone up.
• At least two people on the team can run the workflow without asking you.

If any of the three fails, fix that workflow before you add another. This keeps the department’s practice small enough to review.

Where this stops

Anything on the page that touches employee monitoring, recording, customer data or regulated information needs your HR, legal or security team. This post is general education, not advice on any of those areas.

What to do this week

• Ask each person on the team, in one line each, where they already use AI at work. Those answers are the raw material for the page.
• Pick the two workflows from that list that recur most often and that someone can check quickly.
• Time each one once, as it runs today.
• Draft the page, share it at your next team meeting, and put a date on the calendar six weeks out to measure again.