Your Product Operating Model is Broken (And AI is Making it Obvious)

Why upgrading your product operating model is finally no longer optional.

Our Product Management heroes have preached it for over a decade or even two. Most product leaders feel at least inspired. Many still struggle to transform their orgs, mostly because of organizational dynamics. We see Product Ops roles and discovery practices popping up in more organizations. Promising signs.​

But it’s not enough. If you’re still running a 2015 operating model – voluntary or not – your business isn’t dying slowly anymore. It’s dying at high speed.

Companies that invested early in discovery and outcomes over outputs transformed their ways of working, hiring, and communication. It wasn’t easy to prove the business impact, so it always felt like a “product thing”. Those unable to convince their org – or uninterested – could sweep it under the carpet. (They might struggle to hire top talent or deliver impact, but hiding it was easier.)​​

Now AI changes everything. Those who built the foundation can layer on the automation/AI accelerator – without rethinking processes or mindsets. They finally prove outcomes over outputs pays off: derisking at hyperspeed (recommend to read Christina Wodtke’s “discovery coding” article).

Suddenly, innovation isn’t a big, expensive bet – it’s a challenge teams can tackle.

The gap between those companies and the rest is widening faster than ever. I’m not talking Vibe Coding or Synthetic Personas alone. I’m talking companies rewiring how they work – versus those running a 2015 playbook in 2026. In mature companies, this gap has never felt bigger.

I am convinced: the real risk isn’t AI killing jobs.

The real risk is turning your product into a legacy magnet: attracting legacy customers, repelling ambitious ones, shrinking your market until there’s not much left. The risk your hires are only those who won’t adapt to cut time-to-value – but repeat the past 20 years.​

This isn’t “fancy LinkedIn operating models”. It’s killing your business – no longer slowly, but at high speed.

---

When even the experts feel behind

In late 2025, Andrej Karpathy – one of the most visible AI experts – admitted he’d never felt this far behind as a programmer. He wrote about sensing he could be “10x more powerful” if he only managed to connect everything that has become available in the last year, and that not doing so felt like a “skill issue”. (read Stefan Wolpers article “Agile’s AI-Driven Paradigm Shift”.)​

If even experts in this field feel behind, what does that mean for a mid-sized B2B SaaS company still trying to convince stakeholders to solve user problems instead of building from a wishlist – or discussing if they should “try a bit more discovery” next quarter?​

We don’t need more drama to see the point. The context of product work has changed. Code is getting cheaper; learning is getting more valuable. Prototypes that used to take weeks now take hours. AI can synthesize feedback, generate concepts, and stress-test ideas faster than any human team.​​

In that world, companies that still treat product as a feature factory with a few AI add-ons are not just slower. They are actively repositioning themselves into the “legacy” corner of their market.

---

The legacy magnet effect

Most leaders worry about being disrupted by a bold new competitor.

In reality, many mature companies are being eroded from the inside by something more subtle: they become the default choice for the least ambitious part of their market – the customers who chose them years ago and never got around to switching.

If you’re leading product in such an environment you know the signs:

• Your product still sells, but more often on discounts, bundles, or long contracts than on excitement.

• Your roadmap is full, but mainly with custom requests and edge cases for existing clients.

• Prospects compare you with newer players and start using phrases like “solid, stable, not very modern”.

This is the legacy magnet at work. It doesn’t happen because your teams are lazy or your product is bad. It happens because your operating model has always been out of synch, but you were able to hide it in front of your customers. But now it is out of sync with what modern product work looks like in an AI-enabled world and your customers can see and feel the difference between a company that cares for their user problems and iterates fast.

You still ship. You still improve. But you’re optimizing a model designed for a time when prototyping was expensive, research was slow, code was the main bottleneck.

That world is gone.

---

Two paths for mature companies

Right now, I see two broad paths in mature companies.

Path 1: AI as a bolt-on to a feature factory

Here, AI enters an organization where strategy lives mostly in slides, not in daily decisions; product teams are judged on velocity and output, not outcomes; “agile” exists as theater – rituals without real empowerment or learning.​

In this context, AI is handled like another feature: a chatbot here, an “AI assistant” there, a few internal pilots disconnected from core value creation, and some trainings on “how to use AI for efficiency”.

What happens? If teams start adopting AI in their day-to-day, it amplifies whatever is already there: specs generated faster, designs produced faster, code shipped faster.

But the underlying system is still the same. No clear strategy. No disciplined discovery. No real commercial ownership in product. No appetite to place bigger bets on your roadmap to stay ahead of the curve. No discussions about rethinking your organizational setup and setting up smaller cross-functional teams to bring innovation to life in small iterations.​

Leaders proudly announce “AI features”, while customers still ask: “How exactly does this solve my problem – and why is it better than alternatives?”​

What separates the two paths is how they handle the cost of innovation: one keeps it expensive, the other rewires it to be manageable even in a mature setup.​​

What sets the paths apart is the cost of innovation. Path 1 keeps it sky-high. Path 2 rewires it manageable.​​

Path 2: Builder Teams as the New Operating System

On the second path, product leaders don’t start by preaching “more discovery” to everyone. They start by re-architecting how product gets done. The shift is structural, not cosmetic.

Instead of trying to transform the whole organization in one go, they carve out builder teams inside the existing setup. Small, cross-functional, outcome-owning teams with enough autonomy to work differently: closer to the problem, closer to the data, and with AI baked into their daily workflow – not added as an afterthought.

These teams don’t just run “better ceremonies”. They:

• Own a problem space end-to-end, from insight to impact (like preached for ages).​

• Prototype and ship using AI-native workflows, dramatically reducing cycle time.

• Treat experimentation as the default, not as a side project.

What changes is the economics of innovation. In a traditional setup, big bets are rare and expensive: you have to align half the org, lock in a roadmap, and live with the sunk cost if it fails. In a builder-team model, the same big bet is decomposed into a series of smaller, faster, cheaper bets – run by a team designed to learn quickly and share that learning back into the organization.​

Suddenly, “innovation is expensive” stops being an excuse. Even in a mature environment with legacy systems, these teams can derisk bold ideas in weeks instead of years, because the operating model around them is built for speed, autonomy, and learning – not just for delivery. Over time, their way of working doesn’t stay isolated. It becomes the template for the rest of the org: practices, rituals, and role definitions are rolled out based on what actually worked in the builder teams, not based on a slide deck.

This is what a modern product organization looks like in practice: not a new set of buzzwords, but a deliberate choice to create spaces where big bets become manageable bets – and where AI is used to reduce the cost of being wrong, not just the cost of writing code.

---

AI exposes your product mess more than it fixes it

If your product organization is already struggling with:

• Feature factories dressed up as “empowered teams”.

• Shipping illusions: lots of releases, flat outcomes.

• Roadmaps that read like stakeholder wishlists instead of strategic bets.

AI will not save you. It will expose you.

Because once you competitor can prototype your “big idea” in a day, run a lean test in a week and iterate three times in a month,

…your old excuses (“too slow to validate”, “too expensive to test properly”) evaporate.

The conversation at the decision-making table shifts:

• From “Can we build it?” to “Why didn’t we test this earlier?”

• From “We need more budget to learn” to “Why are we learning so slowly with the budget we already have?”

Companies with a modern product operating model will have good answers. Companies with a legacy model will have reasons.

---

This is a people problem, not a tooling problem

It is tempting to frame this as an AI tools race.

It isn’t.

As Stefan Wolpers points out, AI adoption in agile organizations is fundamentally a people and culture challenge, not a tooling challenge. The same is true for product.​

The critical questions haven’t changed in the past decade – but now it’s finally no longer a “nice to have”:

• Are we willing to change how we decide, not just how we code?

• Are we ready to let teams work in short, evidence-based loops – even if that feels slower at first?

• Do we accept that our product org must own commercial outcomes, not just delivery?​

If the answer is no, AI will mostly decorate a system that was already creaking. If the answer is yes, AI becomes an accelerator for a new way of working.

A different kind of risk statement

So no, this is not another “AI will take your job” story.

For CEOs and product leaders in mature SMBs, the risk sounds more like this:

If you keep your old operating model, you won’t suddenly fall off a cliff.

You will just wake up one day and notice that your best customers have moved on, your most ambitious employees have left, and your roadmap is dominated by small requests from clients too tired to switch.

You’ll still have a product. You’ll still have revenue. But you’ll be a legacy product in a legacy corner of the market.

The good news: the first step out of this trajectory is not a massive AI budget.

It’s an honest question:

“If a prospect labeled us ‘the legacy option’ today, what proof do we have that they’re wrong – in how we work, not just in what we say?” (and yes – it’s happening).​

Answer that with courage, and the rest – tools, skills, AI capabilities – can follow.