At a Glance
AI is dramatically reducing the cost and effort of building products. What was once scarce, the ability to create, is rapidly becoming abundant. This shifts the role of product management from building solutions to making better decisions. The real challenge is no longer execution, but judgement: deciding which opportunities matter, which risks are worth taking, and which paths to ignore. As AI generates more options at every stage of product development, product leaders must strengthen their ability to assess context, evaluate trade-offs, understand commercial impact, and make decisions aligned to strategy. In the AI era, judgement is becoming the most valuable product management capability.
AI just made almost everything buildable, which raises a question for every product team: if anyone can build anything, what is actually scarce now? The answer is judgement, knowing what to build, who it is for, and why it matters.
Consider how quickly the cost of building collapsed. The share of Google’s new code written by AI climbed from about a quarter in late 2024 to roughly half a year later, and Sundar Pichai now puts it at 75 per cent. When a working version of almost anything can be spun up in an afternoon, building stops being the thing that separates good teams from the rest.
So if building is no longer the hard part, then the judgement calls are, and we make them constantly, at every zoom level. What is changing is the sheer volume, because AI keeps handing us more options to weigh at every turn.
There is a deeper reason too. Something that looks like a product is not the same as something that is one, and that gap is exactly where judgement lives. As Kendra Vant says, without a reliability layer it is simply a prototype. Closing that gap, deciding what reliable enough means, what to ship and what to hold, is a judgement call. AI can produce the thing. It cannot decide whether the thing is actually ready.
This article is about those calls, the ones we make every day without naming them, how we get better at making them, and why AI, far from taking our job, may be handing the real one back.
What does judgement mean?
Strip it back and judgement is what we do when we are handed a set of choices and have to call which one to back. We weigh what we know, what the data says, and what we sense about where things are heading, then we commit to the path most likely to deliver the outcome we promised the people who count on us. The dictionary calls it forming an opinion by discerning and comparing, which is fine as far as it goes, but in product the stakes are sharper, because every call also closes off the paths we chose not to take.
Can AI make judgement calls?
You can certainly ask it to, and it will answer. AI is genuinely good at parts of this. It will lay out the options, summarise what the data says, draft the analysis, even argue both sides more patiently than most of us manage on a Friday afternoon. That is real, and we should use it.
But producing the analysis is not the same as making the call. Choosing which problem is even worth solving, weighing a decision against context the model cannot see, the promise you made a customer last quarter, the bet the company is quietly placing, what your brand can and cannot be caught doing, and then carrying the consequences when it goes wrong, none of that delegates cleanly. That part sits with a person, and it always will.
That only gets harder as AI floods us with more to choose from, more insights, more opportunities, more plausible solutions at every turn. A widely cited Pendo analysis found that 80 per cent of features in the average product are rarely or never used. That was the picture when building was slow and expensive. Now that it is fast and cheap, we can build far more of what nobody wants, far more cheaply. The constraint was never building. It was choosing.
So we have to sharpen our judgement.
So how do we sharpen it?
Picture the work as a cascade. A vision at the top, then strategy, then the customer problems worth solving, then the opportunities we invest in and in what order, and finally the solutions we build. At every step the road forks, and the path we choose not to take or to take is a judgement call.
Here is what AI changes about that picture. It does not just hand us the obvious two or three paths at each fork, the ones we would have drawn ourselves. It surfaces the faint paths off to the side, the options we would never have thought of, and it does that at every level at once. Look at the diagram and notice how many more lines there are than the few we bother to name. The volume of paths has exploded, and so has the number of calls we have to make. More to choose from is not the same as easier to choose. In fact, it is the opposite.
So how do we make a good call when the fan of options is this wide? A few things matter more than they ever have.
Good inputs
It starts with the inputs. We have to tell good data from bad, and we cannot get it all from one place. AI is a source, and a useful one, but it is not the source. We still need to sit with our customers, watch what they do rather than what they say, and notice the things they leave out. We need to engage our stakeholders more deeply, not less, because a good part of the answer lives with them, in context that no model has yet seen. Judgement is only ever as good as the inputs feeding it, and the valuable inputs are rarely the easiest to gather.
Commercial instincts
We also need a real feel for the commercials, even if you work in government or the not-for-profit sector. Every path returns something different, and we should be able to say roughly what, where the value comes from, what it costs to get there, and how long before it pays back (in whatever form value takes). A choice that delights customers but does not move the “business” is only half a choice.
Market foresight
Then there is the market, which never sits still. Good judgement reads the way things are shifting, the competitor that just changed the game, the behaviour that is becoming normal, the regulation coming over the hill. It projects forward to where the world will be when we actually ship, not where it is today. We are always choosing for the future, so we have to be able to see a little of it.
Risk management
We have to interrogate the risk. For any path we are drawn to, we pose the harder questions before we commit. What has to be true for this to work? What happens if we are wrong? Who is exposed if it fails? Then we ask whether that level of risk actually fits our appetite, because a brilliant bet the organisation cannot afford to lose is not a brilliant bet.
Strategy shapes our judgement
And through all of it, we keep coming back to the strategy. Every fork gets checked against where the organisation is trying to go and the goals it has set. A path can be popular, technically elegant, and championed by your loudest customer, and still be the wrong one if it pulls us away from the strategy. The strategy is what tells us which of the many paths is actually ours to take.
So, a choice
AI has not lightened the job. If anything, it has intensified it, handing us more paths, more options, more calls to make at every turn. The administrative layer, the tickets, the roadmaps, the status updates, is being automated away. What is left, and what is multiplying, is the judgement.
So we have a choice, and it is worth being honest about both sides. One path is to lean toward the building. Use the time AI frees up to prototype, to ship, to become more of a maker, and over time to move into engineering or delivery work. That is a real and respectable choice, and plenty of good people will make it.
The other is to get closer to customers, weigh the options properly, and become the person who makes the hard choices well. If you stay in product, this is the job now. Not the building, which AI will only get better at, but the judgement, which it cannot do for us. We make these calls every day, mostly without naming them. We have to name them, own them, and get deliberately better at making judgement calls.
