IntellectraAI

Free worksheet · no account or email needed

A practical AI task audit for project managers

Project managers connect people, make trade-offs visible, and help teams decide what comes next. Some supporting tasks may be good candidates for AI assistance. This worksheet helps you test one task safely while keeping human judgment in charge.

Allow about 20 minutes to choose a task and plan a test. Use paper or a local document. The exercise is complete on this page; the assessment at the end is optional.

1. Choose one routine task

List three things you repeat in a typical week: perhaps a status summary, proposed meeting action items, or a check that a risk log is readable. Pick one frequent, low-stakes task that your organization permits you to test.

Fill in one copy for each candidate task

  • Repeated task: ____________________
  • A good result: ____________________
  • Human decision: ____________________
  • Data or policy restriction: ____________________

Keep client names, confidential files, employee details, unreleased plans, and other employer-sensitive information out of public AI tools. If a tool or data category is not approved, use only a fictional example.

2. Separate drafting from judgment

Possible drafting help

Structure, first-pass wording, a checklist of missing fields, and alternative ways to summarize facts you have already verified.

Human decisions

Which facts are true, which risk matters, who needs to be consulted, whether a deadline is realistic, and what to communicate to stakeholders.

A readable status update helps only when the underlying facts and decisions are sound.

3. Test with fictional facts

Imagine a fictional training rollout: a practice session is scheduled for Thursday; a draft guide needs a final accessibility check; and a stakeholder review is waiting for a confirmed time. No real people, client details, or internal files are involved.

If your organization's policy permits the tool, try this prompt with those facts:

Draft a short project status note using only these fictional facts: the practice session is scheduled for Thursday; the draft guide needs a final accessibility check; the stakeholder review time is unconfirmed. Use sections for progress, open decisions, and proposed next actions. Mark any missing owner or due date as “to confirm.” Do not invent completion, risk ratings, or commitments.

Treat the output as a draft. Did it invent an owner, say the accessibility check was finished, or turn an unconfirmed review into a firm deadline? Correct or reject it. Compare the draft with a note you would have written without the tool. If it did not help, record that result too.

4. Decide whether to repeat the test

  1. Did the draft preserve every source fact and flag what is unknown?
  2. Can I verify the output against a source of truth I control?
  3. Does this task contain data that changes which tool or workflow is allowed?
  4. Did drafting help, or did it simply move the work into correcting the draft?
  5. Which part still required my context, relationships, or judgment?

There is no universal pass score. This is a decision about one task in your own setting, within your organization's rules. Note what you tried, what you changed, and what you would do differently next time.

What to do next

If the test helped, repeat it on another approved, low-risk task and keep a human review step. If it did not, try another workflow or strengthen the capability the exercise revealed, such as prioritization or stakeholder alignment.

For a broader profession-specific starting point, our free assessment asks how you work today and returns a readiness profile, capability gaps, a learning path, and a four-week Development Plan. It is educational and does not predict employment outcomes.

Start my free assessment

Free. No account required. Results are available immediately.