Project Management in the Age of AI: What Changes and What Doesn't

Every general-purpose AI tool can now produce a project plan in under a minute: a work breakdown structure, a schedule with dates, a risk register, a RACI matrix — formatted better than most of us would manage by hand. And almost every one of those plans is quietly wrong in the same way: it invents a sequence nobody validated and hands it over already looking finished.

What AI genuinely does well

• First drafts — a WBS, charter skeleton or risk register that is 70% right in seconds beats a blank page every time.
• Formatting and consistency — naming conventions, status report structure, meeting notes turned into action lists.
• Breadth of suggestion — it will list risks you forgot existed, because it has seen more project post-mortems than any of us.
• Translation — re-explaining the same plan for executives, engineers and vendors without you writing it three times.

Where it quietly fails

The failures are not typos. They are confident, plausible, and invisible until delivery:

None of these are reasons not to use AI. They are the reasons the review step is now the most valuable hour in your week.

The judgement that remains yours

1. 1Validate every duration against your team's actuals — a model knows industry averages, not your Tuesday.
2. 2Confirm every dependency with the person who does the work; convenience dependencies invent a false critical path.
3. 3Assign a named owner and a date to every risk and action the model lists. A risk owned by 'the team' is a feeling, not a risk.
4. 4Decide what gets cut when the date moves. AI can rank options; it cannot be accountable for the trade-off.
5. 5Read the float. When a model hands you a forty-line schedule, someone still has to say 'this task has three weeks of float, we can move it.' That sentence is the job now.

Why the fundamentals matter more, not less

When documents were expensive to produce, producing them was most of the visible work. Now that they are free, the differentiators are the skills that were always underneath: critical path analysis, float, earned value, dependency management, a RAID log with named owners. These are exactly the things you need to catch a plausible-but-wrong AI plan before it becomes a plausible-but-wrong project.

A practical working agreement with AI

• Use AI for the first draft, never the final one — every artifact gets a human review pass before it is baselined.
• Treat AI estimates as a challenge to your own, not a replacement: where they disagree, find out why.
• Keep the audit trail human — decisions, change approvals and gate sign-offs need a named person, not a prompt.
• Never paste confidential scope, budgets or personnel issues into a tool your organisation has not approved.

The project managers who thrive in the next decade will not be the ones with the best prompts. They will be the ones who can look at a flawless-looking plan and spot the one assumption that was never true.