A colleague recently asked me a question that I haven’t been able to stop thinking about.

We were talking about the Product Operating Model and the eight pillars we use to describe it: Product Principles, Vision, Strategy, OKRs, Teams, Tools and Rituals, Discovery, Delivery and Go-to-Market.

His challenge was pretty simple: what happens to all of this when AI can essentially do anything?

Not literally anything, at least not yet, but it isn’t difficult to imagine where this is going. The cost and time required to perform many kinds of knowledge work is collapsing quickly. Research that once took weeks can happen in an afternoon. Customer feedback can be continuously analysed. Concepts can be generated almost instantly. Prototypes can turn into working software. Code can be written, tested and deployed with much less human effort, while marketing materials, documentation and analysis are increasingly generated alongside the product.

So what happens to the Product Operating Model if execution is no longer the constraint it once was?

It’s a fair question, and I think it gets at something important about what a Product Operating Model is actually there to do.

What happens when building things gets really cheap?

For most of the history of software development, execution capacity has been scarce. There are only so many engineers, designers and product managers, only so many hours in the quarter, and every meaningful idea has to compete with a long list of other things the organisation could be doing.

A lot of the machinery we associate with product development has grown up around that constraint. We prioritise backlogs, coordinate teams, plan quarters, write requirements, manage dependencies, schedule research and negotiate for engineering capacity because there is never enough time to do everything.

AI is clearly going to compress a lot of this.

Some practices will probably disappear entirely, while others will happen continuously in the background. Activities that once required meetings, documents or even dedicated roles may increasingly become things an agent just takes care of.

I don’t think there is much value in pretending otherwise. If AI can do parts of product work faster and better, we should absolutely let it.

Where I think the argument gets more interesting is when we make the leap from “execution is becoming much easier” to “therefore we need less direction.”

I’m not convinced that follows.

A five-minute software development thought experiment

Imagine that tomorrow morning your organisation gets access to an AI system capable of implementing almost any software change in five minutes.

You describe what you want and, for the sake of the thought experiment, it more or less builds it.

The engineering backlog suddenly stops being much of a constraint.

At first this sounds incredible, and it probably is. Then I suspect things get chaotic fairly quickly.

Sales has dozens of customer requests it would like implemented. Marketing has ideas for acquisition and conversion. Support has a long list of recurring pain points. Leadership has strategic initiatives. Product has hypotheses it wants to test. Competitors have released features you could now copy almost immediately. Meanwhile, AI has analysed your customer data and generated another few hundred perfectly plausible ideas of its own.

The hard question is no longer whether you can build something.

The hard question is whether you should.

That seems like a small distinction, but I think it changes the nature of the Product Operating Model quite significantly.

AI creates abundance, but someone still has to create coherence

We tend to think about operating models partly as systems for organising work. That makes sense in a world where getting the work done is difficult, expensive and slow.

In an AI-native organisation, I think the Product Operating Model increasingly becomes a system for directing a huge amount of available intelligence and execution capacity toward something useful.

Vision matters because all of those decisions still need some shared understanding of the future the organisation is trying to create.

Strategy matters because an organisation capable of pursuing almost anything needs to be very clear about what it intends to pursue, and probably even clearer about what it intends to ignore.

Principles matter because more decisions will be made locally by people and agents without every choice being escalated through management.

Goals matter because an enormous increase in execution capacity doesn’t help much if we aren’t clear about the outcomes we are trying to change.

The other pillars evolve as well.

Discovery may become less of a scheduled activity and more of a persistent sensing system, where AI continuously brings together customer behaviour, qualitative research, market signals and emerging opportunities.

Delivery will almost certainly change substantially as specifications, handoffs, coding, testing and deployment become increasingly compressed.

Go-to-Market changes too. Messaging, enablement, launch materials, segmentation and measurement can increasingly be developed alongside the product rather than being treated as a separate sequence of work afterwards.

To me, this suggests that the pillars themselves don’t necessarily go away. What changes, sometimes quite dramatically, is how the work inside them gets done.

The bottleneck moves

This is perhaps the simplest way I’ve found to think about it.

AI is reducing the cost of execution much faster than it is eliminating the need for judgement.

When software is expensive to create, organisations naturally spend a lot of energy figuring out how to get things built. If software becomes dramatically cheaper to create, the constraint starts moving somewhere else.

The scarce resource becomes good judgement about what is actually worth doing.

Which customers do we care most about? Which problems are genuinely important? What kind of experience are we trying to create? Where can we build an advantage that isn’t immediately reproducible by somebody else with the same AI tools? What are we willing to say no to? What does success actually look like?

Those are questions about direction, choice and intent.

They are also questions that sit right in the middle of principles, vision, strategy and goals.

Ironically, the parts of a Product Operating Model that can sometimes feel furthest removed from the day-to-day act of shipping software may become more important as shipping software itself gets easier.

The operating model probably gets lighter, not smaller

None of this means the Product Operating Model should stay exactly as it is while the world around it changes.

I would expect the opposite.

An AI-native operating model should probably involve less documentation for the sake of documentation, fewer handoffs, less backlog administration, fewer status meetings and much less manual work assembling information that machines can already synthesize.

There is a lot of traditional product process that exists because coordination has historically been difficult. We shouldn’t preserve that process just because we’re used to it.

What I don’t think disappears is the need for the organization to be intentful about where it is going, what it believes, what outcomes matter and how decisions should be made.

In fact, an organisation with a huge amount of autonomous execution capacity probably needs to get much better at communicating that intent.

Here is where we are going. Here are the customers we care most about. Here are the outcomes we want to change. Here are the strategic bets we are making. Here are the principles we want decisions to reflect. Here are the boundaries within which teams and agents can operate independently.

Once that context is clear, you can give the system a lot more freedom to move.

Maybe the question itself is changing

So I’m not sure the most interesting question is whether AI makes the Product Operating Model obsolete.

I think the more useful question is: what is a Product Operating Model for when execution becomes abundant?

My current view is that it becomes less about managing the production of work and more about maintaining coherence across an organisation that is capable of moving much faster than organisations have historically been able to move.

AI gives us enormous leverage, but leverage on its own isn’t particularly useful. Without enough direction, you can just end up producing an enormous amount of activity very efficiently.

The organisations that benefit most from AI may not ultimately be the ones that automate the greatest number of tasks or ship the greatest number of features. They may be the ones that get particularly good at connecting vision, strategy, outcomes, discovery, execution and learning into a system that can move quickly without losing its sense of direction.

I don’t think the Product Operating Model disappears in that world.

I think it evolves with it.

And if the cost of doing things continues to fall, figuring out what is actually worth doing may become one of the most important capabilities an organization can have.