AI MANIFESTO

When execution is cheap, judgement becomes more valuable

AI makes it easier to produce things, but knowing what deserves to be done—and whether the result is useful—takes judgement. The opportunity is to accomplish more while raising the standard of the work.

You can feed a model context. You cannot delegate accountability.

Understand before accelerating

Before anything gets built, someone needs to understand who it is for, what matters to them and which constraints are real.

AI can help explore those questions. It cannot make them unnecessary. When it becomes easier to act on an idea, we need to be more deliberate about which ideas we act on.

Research and alternatives are useful when they help us choose a direction worth pursuing.

Expertise gives speed a direction

Experience helps us notice the missing requirement, the misleading comparison and the exception that breaks an otherwise sensible approach.

It tells us which source to trust, which question to ask and when an answer needs checking.

That knowledge becomes more valuable when we can act on it faster. Making good use of AI requires curiosity, domain knowledge and a willingness to challenge the output.

Use the capacity to raise the standard

Faster execution creates room to test assumptions, explore alternatives and follow the detail through.

To check the relationship between documents. To test the awkward scenario. To notice where a system will fail before someone has to depend on it.

I use AI to help catch contradictions across specifications and develop repeatable test scenarios. The capacity gained should make the work more thorough.

Make the work useful

More documents, tickets or prototypes do not automatically mean progress. Their value depends on whether they help someone decide, build or deliver.

Scattered source material can become a clear brief. An idea can become a prototype people can test. A complicated workflow can become a handover the team can act on.

The measure is what the work makes possible for the people using it.

Stand behind the result

A polished document can rest on a bad assumption. A working prototype can solve the wrong problem. A confident answer can be unsupported.

If I put my name to something, I need to understand it well enough to explain the choices, recognise its limits and correct it when it falls short.

“The model wrote it” is never enough.

Next steps

Put better judgement to work

If this approach sounds useful, let’s talk about what you’re working on.