John Laverick

Idea

As execution becomes cheaper, where do we need clarity and where can we afford uncertainty?

AI reduces the cost of implementation, experimentation and rework. That can lower the amount of certainty needed before trying something, while increasing the value of clarity about the problem, hypothesis, boundaries, learning goal and consequence of being wrong.

Current thinking

AI is making execution cheaper, but I no longer think the consequence is simply that we need to be more certain before we build. In some situations the opposite is true: if we can build enough of an idea in an afternoon to discover that it is wrong, then being wrong has become much cheaper.

That changes what clarity is for.

If a piece of work takes six months and a large investment to deliver, it is rational to spend a lot of time trying to establish that the proposed solution is right before committing. When the cost of testing the proposition collapses, we can afford to act with less certainty about the solution.

What we still need is clarity about the problem, the outcome, the boundaries and what we are trying to learn.

We can be uncertain about the solution while being very clear about the experiment.

That distinction matters because building the wrong thing quickly is not always a failure. Sometimes it is the mechanism through which we discover what the right thing is.

We might believe users will behave in a particular way and discover that they do not. We might automate a process and realise the real problem sits upstream. We might build a feature and learn that nobody values it enough to use it. If those discoveries take months, they are expensive. If they take hours or days, they are useful information.

So cheaper execution creates two effects at once.

It increases the value of clarity because AI can act on vague direction with extraordinary speed. But it also reduces the amount of certainty we need before trying something, because changing direction is cheaper.

The useful question is therefore not simply “Are we clear enough to build?”

It is also: “What are we building for?”

Are we building the answer because we have already made the decision? Or are we building something deliberately to find the answer?

Those are different activities and they should be governed differently.

When we are experimenting, clarity means knowing the hypothesis, the limits, the consequence of being wrong and what evidence would cause us to change direction. When we are operating something people depend on, clarity also includes ownership, support, assurance and the standard the service must meet.

AI has not made ambiguity harmless. But it has made some mistakes dramatically cheaper to discover.

That is a better version of the idea for me: as execution gets cheaper, the premium moves from certainty about the solution towards clarity about intent, learning and consequence.

Writing on this idea