The situation
The client, a specialist consultancy working on a major infrastructure programme, had project managers presenting to sponsors, boards and delivery teams every week. AI tools were already in everyone’s hands, and the packs were arriving faster than ever. Landing, though, is a different thing from arriving: titles that described topics instead of making points, ten ideas per slide, and AI-generated content passed on without an owner.
That last habit now has a research name, “workslop”: output that looks like work but adds none. Nobody wants to receive it, and nobody enjoys discovering they have sent it.
The approach
One hour, protected question time, and a rule set out at the start: the story comes first, the slides prove it, and AI never gets the final say.
Frameworks before tools
Audience first, one message per slide, titles as messages, the Pyramid Principle and SCQA, taught with worked examples.
A quiz with teeth
Three interactive questions built on research under twelve months old, ending on the statistic that pivots the whole session.
The 80/20 rule
AI accelerates the build; the steer and the sign-off stay human. Four things AI will not do for you, stated plainly.
Self-serve from day one
Everything lives in a hub the team keeps: the deck, a full playbook, and a prompt library with copy buttons and links to their AI tools.
What was delivered
The deck runs in any browser with keyboard, touch and fullscreen support. The playbook is a working reference with a searchable, filterable prompt library. Every statistic carries a named source. The whole thing ships as static files: no logins, no installs, nothing to maintain.
The proof is in the product: the toolkit that teaches the workflow was built with the workflow.
Why it matters on a Tuesday
The session was the launch day. The hub is for every day after: the moment a PM faces a blank page with a steering committee looming, the prompts are one click away, the frameworks are one page away, and the standard for “good” has already been agreed in the room.
- Prompts copy to the clipboard and open straight into Copilot, Claude, Gemini, ChatGPT or Perplexity
- The playbook reads in fifteen minutes and stays open while packs get built
- The quality bar, render it, review it, own it, is written down where everyone can point at it
Judge the output yourself
The full hub is live: the deck, the playbook and the prompt library, exactly as the team received them.