The Pocket Test

There is an old definition of a product that I find hard to forget.

A product is something you can put in your pocket and walk away with.

It is slightly absurd. Which is probably why it works.

It does not try to be complete. It tries to be useful. And it is useful because it forces one question before anything else: when this conversation ends, what remains?

Then AI arrived. And the pocket became too small.

The problem is not that AI is complicated.

The problem is that very different things now travel under the same name.

Training a team to use AI better. Building reusable prompt libraries. Creating software with AI assistance. Building software whose value depends on AI at runtime. Delivering a service faster because AI is inside the engine.

These are not variations of one thing. They carry different costs, different timelines, and different definitions of what done means.

When they collapse into one label, something is lost that is difficult to recover later.

There is a specific kind of confusion that happens when categories collapse.

It is not felt equally by everyone in the room.

The person closest to delivery can see it early. They know which category is being discussed and which one is actually being built. They can feel the distance between the expectation and the reality taking shape.

The person thinking about possibility often cannot see it yet. Not because they are wrong. Because they are working from a different level of resolution.

This asymmetry is where the real cost lives.

Because the confusion does not stay in the room where it started. It moves forward. Into proposals. Into what the scope covers. Into what done means. And eventually, into delivery.

The risk does not disappear when the words are agreed on.

It moves to whoever can see the difference.

What AI needs now is what that old definition provided then.

Not a bigger label.

A better question. Asked early, before anyone has committed to anything.

Not: does this involve AI?

But: where does the intelligence actually sit, and what does the client continue to receive after the engagement ends?

That question has different answers. Sometimes the answer is capability inside people. Sometimes a reusable asset. Sometimes software. Sometimes a platform. Sometimes an ongoing service where AI is the engine but the delivery is still human-shaped.

Each answer leads to a different conversation. Different expectations. Different outcomes.

The question comes first.

The map comes second.

First, make the category visible. Then profile the shape of the solution.

That is what the diagram below is for.

The pocket test was almost too simple to take seriously.

But it did one thing well.

It made the category visible before anyone had committed to anything.

In AI, that moment still exists.

It just requires a different question.