Building it is no longer the hard part.

Production got cheaper. Attention did not.

A small team can now produce a plausible version of almost anything in a week: the screens, the copy, the landing page, a working slice of the product. What has not changed is how many things anyone is willing to try, learn, trust or pay for. The constraint moved, and it moved to the part that cannot be produced.

A team measured on output will produce output

Ask for more shipped and you will get more shipped. That is the uncomfortable part: output is the easiest thing to measure and now the easiest thing to increase, so it is the thing that gets rewarded. Nobody is waiting for more of it.

The cost is not only the wasted week. Every additional surface is another thing to maintain, explain and keep consistent with everything else. A team that doubles what it produces without changing how it decides does not get twice the product. It gets the same product with more edges.

Speed is only worth what it teaches you

The honest question about a faster team is not how much it built. It is how much it learned per attempt.

That turns throughput into something useful. Before building the next version, write down the question it answers and the result that would change your mind. If no result would change anything, you are not running an experiment — you are producing. Both are legitimate, but only one of them earns the speed.

This is also where AI helps most and is claimed least. It is very good at producing the variant, the synthetic case, the edge input, the version of the flow you would not have had time to try. It cannot tell you which answer mattered.

"Can we build it?" is not the hard question any more

For most software, the answer is now yes, at a price that keeps falling. The question that decides whether the work was worth doing is whether anyone will choose it — over the tool they already have, over the workaround they have grown used to, over doing nothing at all.

That question does not get answered by building more carefully. It gets answered by a promise someone recognises, an experience that delivers it, and evidence from people who are not on your team.

The advantage is a shorter distance

In a year, the teams that are ahead will not be the ones with more people in the room. They will be the ones where the distance between what they learn and what they are allowed to change is short.

That distance is usually structural, not intellectual. A team learns something real in week six — and can only change the screen, because the positioning is signed off, the brand is delivered and invoiced, and the build is running to a spec agreed in month one. The learning was good. The system around it could not absorb it.

When strategy, brand, product and engineering are part of the same work, that same insight can change the opportunity, the promise, the interaction, or what the system actually does. Whichever one is wrong.

What to do this month

• Name the metric your team is actually rewarded on. If it counts output, expect output.
• For the next thing you build, write the question it answers and the result that would change your mind — before it is built.
• Find your longest distance: the last real insight you had, and the thing it was allowed to change.
• Ask, about the thing you are building now, who would choose it, and over what.

AI expands what a team can make. Judgment decides what is worth making. Working out which opportunity deserves the build is where our Product & Strategy work starts, and AI & Agentic Systems is how it gets built once that is settled.

Where this leadsProduct & Strategy