It does whatever you ask. It rarely challenges you. It rarely pushes back. It rarely says: “This is a bad idea.” And if it goes off script, well, that’s probably because your instructions were too vague.
Finally, executives got what they always dreamed about: a builder who simply builds.
So why are the results still often subpar?
Because great products do not come from obedience.
Great products come from creative tension. They come from people with different skills, backgrounds, instincts and opinions arguing their way towards a better solution. Not arguing for the sake of arguing. Arguing because the customer problem is not obvious, the solution is not obvious, and the first idea is rarely the best one.
You need engineers who challenge feasibility and suggest better technical paths.
You need designers who challenge assumptions about usability, behaviour, accessibility and clarity.
You need product managers who challenge whether the problem is worth solving in the first place.
And you need all of them working together, not merely executing orders.
But many organisations do not work this way.
They use a backwards model.
They do have engineers, designers and product managers. On paper, they have cross-functional teams. In practice, those people are often not empowered to use their full talent.
Engineers are asked to write code.
Designers are asked to make things look nice.
Product managers are asked to gather requirements, manage stakeholders and report, report, report.
Marty Cagan has been banging the drum about this wastefulness for years. Decades, even. And still, too many executives continue to underuse the people under their leadership.
The result is predictable. Talented people become frustrated because they are not allowed to do the work properly. Organisations move slower than they should. Products become weaker than they should. And leadership keeps wondering why execution does not magically turn into innovation.
So what changes with AI agents?
Put simply: AI agents expose the biggest failure of many executives - their inability to empower people.This can happen in a few ways.
An executive might use AI agents to build their ideas exactly as imagined. No pushback. No challenge. No uncomfortable conversation. Just execution.
And then the result lands flat.
With a little self-reflection, they might finally understand why they once had a product manager whose job was not to “manage requirements”, but to validate ideas and separate good bets from bad ones.
Or take engineering.
AI agents can write code faster than any human. They can generate features every day. They can produce an impressive amount of output.
But if those features are not adopted, not valued, not used, and not loved by customers, then maybe the organisation will finally ask: is there something more to engineering than writing code?
The answer, of course, is yes.
Good engineers do not just implement solutions. They shape them. They reduce complexity. They see risks others miss. They understand trade-offs. They know when a requested feature is really a symptom of a deeper product or system problem.
Or take design.
Executives love having opinions about design. And AI agents are increasingly good at turning those opinions into pretty pictures.
The problem is that pretty is not the same as useful.
Products start to look the same. Usability suffers. Accessibility suffers. Consistency suffers. The user experience becomes a collection of surface-level decisions rather than a coherent system.
So no, design was never just “making it look nice”.
AI is not a solution for a broken culture
In fact, AI may make broken cultures more visible.For years, many organisations have used only a fraction of the value of their engineers, designers and product managers. They hired talented people, then reduced them to narrow execution roles.
Now agents make that failure painfully obvious.
If all you ever wanted from your teams was obedience, AI can give you more of it than ever before.
But obedience does not create great products.
Judgement does. Collaboration does. Pushback does. Discovery does. Taste does. Care does.
The question is how executives will react when they see the results.
Most will probably blame AI, just like they previously blamed people who did not automatically agree with their every “brilliant” idea.
But maybe some will pause.
Maybe some will realise that the problem was never the speed of execution. Maybe they will finally understand that the real bottleneck was poor product judgement, weak empowerment and a culture where people were expected to follow instructions rather than do their best work.
Maybe they will finally listen to Marty Cagan and give their talented engineers, designers and product managers the chance to work properly.
But maybe, more likely, this is just another beautiful dream.