AI for Project Management

Most projects start as a fuzzy sentence: "We need to move the newsletter to a new platform by Q3." A project brief turns that into a shared starting point — goal, scope, who's involved, and what "done" looks like. The trick is to give the AI your messy raw material and let it impose the structure. Don't write the brief yourself first.

> Act as an experienced project manager. Turn my loose description
> into a one-page project brief.
>
> The project: moving our 4,000-subscriber monthly newsletter off our
> current email tool onto a new platform because the old one mangles
> templates. Live by end of Q3. Involves me, our designer, and IT.
> Small budget.
>
> Use these headings: Objective, Scope (in/out), Stakeholders,
> Deliverables, Success Criteria, Assumptions, Open Questions. For
> anything I haven't told you, put it under Open Questions, don't guess.

That last instruction matters most. Without it, AI fills gaps with invention; with it, the unknowns return to you as a checklist.

Once the brief is agreed, you need a plan: phases, tasks, owners, and rough timing. This is where AI saves the most time — it's genuinely good at breaking one big thing into its parts.

> Based on the brief above, break the project into phases. For each,
> list the tasks, suggest an owner (project lead, designer, or IT),
> and estimate effort in days.
>
> Present as a table: Phase, Task, Owner, Effort, Dependencies. Order
> so dependencies come first. Flag any task where the owner is unclear.

You'll get a tidy, dependency-ordered table to paste straight into a spreadsheet. Now turn it into a timeline — give the AI your real constraints and let it do the date arithmetic:

> Turn that task list into a week-by-week timeline starting Monday
> 6 July. Assume each person gives this 2 days a week, not full-time.
> Respect the dependencies. Show which week each task starts and
> finishes, plus the projected completion date.

A risk register lists what could go wrong, how likely and damaging each is, and your response. Most people skip it because it feels like homework. AI removes that excuse, drafting a thorough one in seconds — including risks you'd forget.

> Create a risk register for the newsletter migration. Think broadly:
> technical, people, timeline, and reputational risks (e.g.
> subscribers getting duplicate emails during the switch).
>
> For each: a short Risk, Likelihood (Low/Med/High), Impact
> (Low/Med/High), and a concrete Mitigation. Sort highest combined
> risk first. Aim for 8 to 10 risks.

A typical result rates each risk and pairs it with a concrete mitigation — for example, "subscribers get duplicate emails during the switch" (Medium / High), mitigated by pausing sends and verifying the import count before go-live. The value is breadth: the AI surfaces risks a tired project lead forgets at 5pm on a Friday.

Status reports eat hours and nobody enjoys them. AI is ideal here because the raw material — what got done, what's stuck, what's next — already lives in your notes. Paste them in and let the AI shape it:

> Turn my rough notes into a clean status report with these sections:
> Overall Status (Green/Amber/Red), Progress, Blockers, Next Week,
> Decisions Needed. Half a page, plain confident language an
> executive can skim in 30 seconds.
>
> Notes: template 80% done, designer off sick 2 days so behind. IT
> exported the list, count looks right. Still waiting on the contract
> signature from finance — now blocking go-live. Test send next week
> if the contract lands.

The same notes serve different audiences with one line of framing. Tell an executive to "give five bullets, leading with the decision you need." Tell the team to "include owners and due dates." Tell wider stakeholders to "keep it friendly and focus on progress."

Different stakeholders need the same news told differently — a routine note to your team needs reassuring framing for a nervous client. AI is excellent at this. Draft once, then re-aim it:

> Go-live is slipping two weeks because a contract signature is
> delayed in finance. Write three short messages announcing this:
>
> 1. Chat message to my team — casual, just the facts and the new date.
> 2. Email to my manager — professional, owns the slip, says what
>    I'm doing to recover it.
> 3. Note to the marketing director — reassuring, quality won't
>    suffer, revised date stated confidently.

One input, three audience-appropriate outputs. To make them sound like you, feed the AI a couple of emails you've written and say "match this voice" — far better than describing your style.

AI accelerates the structure and writing of project management. It cannot do the judgement. The AI drafts briefs, plans, and reports, lists tasks and dependencies, and re-tones messages. You own whether the plan is realistic, the priorities and trade-offs, the real likelihood of each risk, and the relationships and commitments behind it.

A useful self-check before acting on any AI output: would you defend it as your own work in a meeting? You can even recruit the AI to find weak spots:

> Before I share this plan, play devil's advocate. What assumptions
> are most likely to be wrong, and what would I regret not questioning
> if this project goes sideways?

That turns the AI from a yes-machine into a thinking partner — but you still make the call. The goal isn't to remove yourself from project management; it's to spend your hours on judgement and relationships, not formatting.