Stop Re-Briefing AI From Scratch Every Monday

AI skills for project managers are encoded, reusable workflows that run from the same instructions each time, without you re-explaining them. That’s the step up from writing good prompts. If you’re re-explaining your process every time you open a new chat, that’s the equivalent of briefing your team from scratch every Monday. Moving one rung up, from prompts to skills, is the change that actually compounds.

If that sounds small, it isn’t. It’s the difference between AI use tied to your memory for prompt wording, and AI use that compounds.

The Maturity Ladder: Where Are You Actually Sitting?

Dharmesh Shah’s maturity model for AI use, what he calls the harness hierarchy, is one of the clearer ways to think about this, and it maps onto PM work better than most frameworks borrowed from engineering.

At the bottom, rung one is raw prompts: you type what you need, every time, from scratch. Rung two is custom instructions or a saved prompt library: better, because you’re not starting blind, but you’re still doing the assembly work each session. Rung three is skills: encoded, reusable procedures built once, so you’re not re-explaining the how every time. Above that sits orchestration and agent networks, a conversation for later.

You can be using AI well and still be sitting at rung one or two. A good prompt library is genuinely useful, real progress over starting cold, but you’re still doing the assembly work every session. Rung three is where something different starts happening.

What Actually Changes When You Build an AI Skill

A skill is a PM workflow captured as a structured instruction set: the kind of thing you’d write up to train a new team member on how you do this here, except AI works from the same instructions every time, instead of whatever you happen to type that day.

Think about the workflows you do on a recurring basis. Status reporting: updates in the format and tone your stakeholders expect. Risk review synthesis: turning a messy conversation into structured register entries. RAID log updates: meeting notes converted into the right categories at the right level of detail. Meeting follow-up: action items, owners, deadlines, in the format your team actually uses. Stakeholder comms translation: technical detail reframed for an executive audience.

Each of these may be something you currently do by writing a prompt, getting an output, and adjusting it slightly differently every time. Encoded as a skill, it’s the same input quality every time, because the how no longer rides on your memory of how you phrased it last Tuesday.

Why This Compounds and Prompts Don’t

Here’s what makes this worth the investment: a skill is an asset. A prompt is a one-off.

A good prompt helps you once, today. Next time, you might remember to reuse it, or use a slightly different version without noticing the drift. A skill, once built, is consistent (the same instructions run every time, not dependent on how well-briefed you were that morning), transferable (someone else on the team starts from the same brief you would), and improvable (fix a gap once and every future use benefits, instead of re-discovering the same fix each time).

That third point is the compounding one. Prompts don’t get better on their own. Skills do, because the improvement lives in the asset, not in your memory of what worked last time.

Map Your Own Rungs

This takes ten minutes with your task list, and a simple way to record what you find.

Here’s what that looks like against a real week. Say your most repeated task is the Friday status update you send to five stakeholders: a project sponsor, two workstream leads, a delivery partner, and your own manager. Right now you open a blank chat, paste in the week’s notes, and type something like “turn this into a status update, RAG-rating each stream.” That’s rung one. The brief gets rebuilt from scratch every week, and the output only matches last week’s if you remember the phrasing.

Run your own most-repeated tasks through the same four questions. Here’s the Friday update filled in as a worked example:

Now answer the same four lines for your own two or three most-repeated tasks. Whichever one lands on “weekly” or “daily” next to rung one or two is your best candidate for rung three. Not every task earns the effort. A task you do twice a year doesn’t need a skill. A task you do every Monday does.

Climbing from rung one to rung three on your two or three most-repeated workflows will do more for your AI output than almost anything else you could spend the same hour on.

Frequently Asked Questions

What’s the difference between a prompt library and an AI skill? A prompt library is a saved collection of prompts you still select and adapt each time: you’re doing the assembly work every session. An AI skill is an encoded, reusable workflow you don’t have to re-explain each time, and it can be improved once to benefit every future use.

Which PM workflows are best suited to becoming AI skills? The workflows you do most often and currently re-prompt for each time: typically status reporting, risk review synthesis, RAID log updates, meeting follow-ups, and stakeholder communication translation. Frequency is what matters: the more often you’d re-brief it, the more worth encoding.

Do I need technical skills to build an AI skill as a PM? No. An AI skill is essentially a well-documented standard operating procedure written for AI to follow consistently. If you can write down how you want a recurring task done, step by step, you have the core of a skill. Ask your AI tool to help you build it.

What’s the task you re-explain to AI most often, and have you ever actually written down how you want it done, once, properly?

Yes - AI helped me to write this :)

Unsubscribe