AI makes building cheaper. It doesn’t make product decisions easier.
As AI changes how products are built, researched and explored, understanding what counts as evidence becomes more important.
AI is changing how digital products are made.
Teams can use it to explore ideas, generate concepts, build prototypes and produce working software faster than before. But it is also changing many of the workflows around product development.
AI can help prepare research, analyse information, identify patterns, synthesise findings and generate possible solutions. I use some of these capabilities myself.
That creates a genuine advantage. Ideas can become tangible earlier. Teams can explore more possibilities and test them before committing significant development capacity.
But it also introduces an important distinction.
As it becomes easier to generate possible answers, we need to be increasingly clear about where the evidence supporting those answers actually came from.
AI is changing more than development
Consider some of the artefacts that increasingly appear during product work.
A persona generated from a prompt may be useful for exploring a possible customer, but it is still a hypothesis about that person.
A journey generated from assumptions may help a team think through a workflow, but it remains a hypothesis about how that workflow actually happens.
A prototype can make a possible solution tangible, but it does not demonstrate that people will adopt it, trust it or pay for it.
AI-generated synthesis can help identify patterns quickly, but its usefulness still depends on the material being synthesised.
This distinction matters because AI can participate in producing and interpreting evidence. It should not automatically become the source of that evidence.
Where did the evidence come from?
If AI analyses twenty customer interviews, the underlying evidence still came from those customers.
If it identifies a pattern in product analytics, the evidence came from actual product behaviour.
If it generates a prototype, the prototype gives us something to investigate. It is not, by itself, evidence that the proposed product will work in the real world.
And if AI generates a plausible description of a customer, market or behaviour from a prompt, plausibility should not be confused with evidence.
The source matters
Useful evidence can already exist in many places: previous research, customer conversations, product behaviour, analytics, support interactions, sales activity, operational information and commercial performance.
Different sources answer different questions.
A customer interview may help explain why something happens. Observation may reveal behaviour that somebody would struggle to describe. Product data can show what happens at scale. Commercial behaviour can tell us something about willingness to commit.
The important discipline is to distinguish between what the evidence actually establishes, what we are inferring from it and what remains assumed.
That becomes more important, not less, when AI makes it possible to analyse and recombine information at speed.
More possibilities still require choices
For a long time, one of the constraints on product development was simply the cost of making something real enough to evaluate.
AI is reducing that constraint.
That means teams can explore more possibilities before committing heavily to one of them. But more possibilities still require choices about where to invest.
The question begins to move from:
That does not mean conducting extensive research around every product decision.
A small, reversible decision may require very little additional evidence. A significant investment, a difficult-to-reverse change or a decision resting on important assumptions deserves more scrutiny.
Product Readiness is about making that investigation proportionate to the decision.
The opportunity is faster learning
The most interesting consequence of AI may not simply be that teams can build the same products faster.
It is that they can explore, test and learn earlier, while changing direction is still relatively inexpensive.
AI can help create the possibilities.
Research, behaviour, product data, operational reality and commercial outcomes can help establish which possibilities deserve greater confidence.
The job is not to remove uncertainty entirely.
It is to understand what we know, what we are assuming, what evidence matters and what that evidence supports doing next.