AI That Actually Works Is Live. Here’s the Launch Conversation.
The book is out. AI That Actually Works is now available, and to mark the launch I went live with David Postlethwaite to talk through the idea at the heart of it: most AI adoption doesn’t fail because of the technology. It fails because of the prompts.
That has been the pattern in nearly every engagement I’ve worked on. Copilot is switched on, the licences are paid for, the pilot group is keen. Six weeks later usage is flat and someone in leadership is quietly wondering whether the whole thing was oversold. It usually wasn’t the tool. People typed a vague request, got a vague answer, decided the technology was mediocre, and went back to the old way of working. That is the single most common reason AI rollouts stall, and it is entirely fixable.
The full launch stream is below. We will be live streaming from 2pm UK time on Friday 19th June 2026
What the book, and the stream, are about
The argument is a simple one. The gap between a team that gets real value from AI and a team that gives up on it is rarely about model capability. It is about whether people know how to ask. Structure the request properly and the same tool that produced a shrug last month starts producing work you would actually send to a client.
That structure is what prompt frameworks give you. They are not clever tricks. They are repeatable patterns that make sure the model has the role, the context, the constraints and the output format it needs to do the job well, every time, regardless of who is typing.
The frameworks
In the stream, and in much more depth in the book, I work through several frameworks. Each one earns its place by being suited to a particular job.
- RTF and APE for quick, everyday tasks where you just need a clean result fast.
- BAB when you are writing to persuade and need a clear before-and-after.
- CARE and CRIT for discovery and analysis, where you want the model to interrogate a problem rather than rush to an answer.
- RISE for structured, multi-step outputs that have to hang together.
- CO-STAR for production-grade prompts you intend to reuse across a team.
- RODES for tightly scoped, well-defined deliverables.
The point is not to memorise all eight. It is to stop treating prompting as guesswork and start choosing the right pattern for the job in front of you.
Why this matters commercially
If you are rolling out Copilot or any other AI tool across a team, prompting is the difference between a licence cost and a productivity return. The technology is largely a solved problem. The behaviour change is not. Give people a small, reliable set of frameworks and you turn scattered, inconsistent use into something you can actually measure and build on.
That is the whole premise of the book. AI That Actually Works covers all eight frameworks in detail, along with a prompt maturity model, guidance on building a shared prompt library, adoption metrics, and a ninety-day rollout plan you can put straight to work.
Get the book, or start for free
Watch the launch conversation above, then pick your starting point. If you want the full system, including the rollout playbook and the shared-library guidance, the book is the place to go: AI That Actually Works on Amazon.
If you would rather try the frameworks before committing to anything, I have put together a free download to get you started straight away. Grab it here: free prompt frameworks download.
Useful Links
AI Enablement for Enterprise: Lessons From a 3,000-Person Week
0 Comments