AI for Project Managers (2026): The Six Artifacts to Hand It, What You Still Check, and a One-Week Plan

How should a project manager use AI?

Use AI for the documents you rebuild every week, not for the decisions. A project manager's week is full of artifacts that turn existing information into a new format: the status report, the RAID log, meeting actions, a schedule risk read, stakeholder updates and change requests. Hand AI the raw inputs, let it draft, and check each draft against a short written list before it goes out. As of 2026 that's the whole practice. The trade-offs, the escalations and the awkward phone calls stay yours.

Most writing on this topic is a list of benefits ("predictive analytics," "smarter resource allocation"). That's not what changes your Tuesday. What changes it is knowing exactly which document to hand over, what to feed it, and what to check before your sponsor reads it. That's this page.

Is this different from AI for managers?

Yes. A people manager's AI question is about leading a team whose work is changing: which tasks move, what quality means now, how roles get re-scoped. I covered that in AI for managers. A project manager's question is narrower and more mechanical: you own a set of recurring documents that pull from tickets, notes and plans, and most of the time spent on them is assembly, not thinking. Before this work I ran product at a national sports platform, and the assembly is the part of the job nobody misses.

The six PM artifacts, and what AI does with each

Here's the map. For each artifact: what you feed it, what it drafts, and the check you do before anyone else sees it.

Notice the pattern. AI is good at the middle column and dangerous in the last one. Every check is the same kind of check: does this match the source, and did a person agree to it?

Three prompts you can paste today

These are the three artifacts that eat the most weekly time for most PMs. Each prompt has the check built into the ask, so the draft flags what it isn't sure of instead of smoothing it over.

1. The status report

Here are this week's closed tickets, open tickets and last week's status report. Draft this week's report in the same format: Done, Next, Blocked, Risks. Use only facts from what I pasted. If a date or a percentage isn't in the source, write [CHECK] instead of guessing. List anything that changed status since last week at the top.

2. The RAID log update

Here is our current RAID log and this week's meeting notes. Propose new entries only (risks, assumptions, issues, dependencies), one line each, with a suggested owner from the people named in the notes. Don't rate probability or impact. Mark any entry where the notes are ambiguous about who owns it.

3. Meeting notes into actions

Here is the transcript. Give me three lists: decisions made (only things someone explicitly agreed to), action items (owner, due date if stated, otherwise [NO DATE]), and open questions. Leave out anything that was only discussed.

Save your project context (the plan, the stakeholder list, the current RAID log) in one saved project in your assistant so every draft starts from your project instead of from zero. Re-explaining the project each session is the reason most PMs quit after two weeks.

Run a one-week time audit before you buy anything

Don't start with a tool. Start with a week. For five working days, write down how long each artifact actually takes and how much of that time is assembly (copying, reformatting, chasing the number) versus judgment (deciding what to say). Assembly is what AI takes. Judgment is what it doesn't.

Then do the arithmetic on your own numbers: assembly minutes per week, times 48 working weeks, times the number of PMs. An example, so the math is clear (these are placeholder numbers, not a measurement): if each of 5 PMs logs 120 minutes of assembly a week, that's 5 × 120 × 48 = 28,800 minutes, or 480 hours a year of copying and reformatting. Your number is the ceiling on what AI can give back. If it's small, you don't have an AI problem, and that's a useful thing to know before you buy seats.

Which AI tools should a project manager use?

One rule: use what's already inside the tools your projects live in before you add anything. Two places to look:

• Your project tool's own AI. Most project tools (Jira, Asana, monday.com, ClickUp, Notion) now offer AI features, often tied to specific plans or sold as an add-on. Check what your current plan includes on the vendor's pricing page before you buy anything new. I'm not quoting prices here on purpose: these pages change often, and a number I didn't read today isn't one I'll print.
• Your company's approved assistant. If the company already pays for ChatGPT, Claude, Copilot or Gemini on a business plan, that's your drafting tool. It's where the saved project and the three prompts above live. If you're the one choosing for the team, the Claude Team plan breakdown covers what a team seat actually buys.

What to skip, at least for the first quarter: separate AI apps that promise to "manage the project for you." Most are a general model with a PM template on top. You'll write a better template, because yours knows your sponsors.

What a PM should never paste in

Project data is usually other people's data. Client documents under NDA, contract terms, rate cards, personal details about team members and anything with a password in it don't go into a personal AI account, ever, and some of it doesn't go into any AI tool without an explicit yes. If your company hasn't written that line down, a one-page AI acceptable use policy does it in an afternoon. Ask for one before you build the habit, not after.

Do project managers need an AI course?

Course pages rank well for this topic, and some are good. My roundup of the best Claude courses for managers includes a Coursera course built specifically for project managers and PMO staff, plus Anthropic's free introductory courses. If you just want to learn the tool, start with a free one.

The test for any paid course: do you leave with one of your own recurring artifacts running? The Shop, my course, is built that way. Each of its 8 lessons ends with a saved piece of your setup, and one of them is a draft-and-approve workflow you run for a week, which for a PM is the status report. It's $397 once, about 4 hours of reading, with a 30-day money-back guarantee, and Lesson 1 is free with no sign-in. It's built on Claude: Lessons 1 to 5 work on Claude's free plan, and Lesson 6 onward uses Claude Team.

Who can help a PMO roll out AI?

I help teams of roughly 10 to 200 people get AI doing real work, and a PMO is one of the easiest places to start, because the artifacts are already standardized. For a group of project managers learning together, the simplest path is The Shop's team seats: $397 for the first seat plus $97 for each teammate, from 2 to 25 people. Five PMs come to $785, ten to $1,270. That's the core of what I mean by AI training for employees: people working on their own documents, not watching a demo. If the question is bigger (which tool the whole company standardizes on, what the data rules are, which teams go first), that starts with my AI Readiness Assessment, $2,500 flat with a written roadmap in two weeks. Or book a free 30-minute strategy call and we'll sort out which of those you need, if either.

What I'd do this week

Monday, start the time audit. Wednesday, run the status report prompt on last week's tickets and compare its draft to the report you actually sent. Friday, fill in the totals. If assembly is more than an hour a week, pick the artifact with the most of it and make the AI draft your default for a month, with the check from the table taped to your monitor.

Sources

• Search demand: DataForSEO keyword overview and AI search volume, United States, English (read October 8, 2026): "ai for project managers" 3,600 Google searches a month, keyword difficulty 21; 428 AI-assistant searches in September 2026.
• The Shop and team seat prices: justinmckelvey.com/academy and /ai-training, as published on October 8, 2026.
• No vendor prices are quoted on this page. Check each project tool's and assistant's own pricing page on the day you buy.