What 20 European Product Leaders Really Think About AI

The serious research on AI and product management is overwhelmingly American. Most of the studies, benchmarks, and survey data that shape how the field understands itself are drawn from US teams, even though the European reality often diverges from it. Europe is not short of voices on LinkedIn sharing their thoughts, but what it lacks is a structured picture of its own. And over all of it sits the noise, which tends to promote more extreme views: the product manager is about to be automated away, every roadmap should be rebuilt around agents, and anyone not shipping AI is already behind.

We wanted a clearer picture of how this looks from inside European product teams, with the volume turned down. So we chose depth over breadth: no panels, no percentages, just twenty long conversations with people doing the work and describing how AI is actually showing up in their products, their teams, and their roles. What follows is what they told us. It is less dramatic than the discourse and considerably more useful.

‍

In Short

• Most customer-facing AI is prioritized to satisfy a CEO, a sales team, or a competitor's release. It often wins deals and then goes largely unused.
• The value that actually lands tends to be unglamorous: internal automation and efficiency, not the flagship feature. Very few leaders measure it with any rigor.
• Trust, not capability, is the binding constraint on adoption. Hallucination and the need to validate everything come up almost universally.
• The dominant framing is that AI is a tool, not magic, and rarely the best solution to a given problem. Several leaders are openly skeptical of building "AI for the sake of AI."
• Productivity gains are real but concentrated in a few places: ideation, research, prototyping, and communication. They are felt rather than measured.
• The human core of the role (judgment, facilitation, accountability, customer empathy) is seen as durable. Whether "product manager" survives as a distinct title is genuinely contested.
• A recurring worry is the apprenticeship: as AI absorbs the entry-level work, leaders fear the next generation is losing the path to senior judgment.

‍

How We Did This

This is qualitative research by design. We were not trying to produce a figure like "sixty percent of product leaders are doing X." A number tells you how many people are doing something, but it rarely tells you why, and the why is where the decisions get made. So instead of a panel and a forced-choice survey, we ran twenty long, semi-structured conversations with people doing the work. That let us capture their reasoning, their contradictions, and the places they flatly disagree with one another, the context a percentage tends to average away.

The interviews were with European product leaders, spanning heads of product, VPs, directors, and senior product leaders across a wide range of sectors and company stages, all working from one shared set of questions so the conversations stayed comparable. We have kept the findings anonymized, identifying contributors only by role and sector, and quotes are lightly edited for readability.

The choice has boundaries, and we would rather state them plainly than dress them up. This is a small, self-selecting, European sample. The answers are self-reported, and nothing here is statistically representative. It is built to deepen understanding, not to size a market.

‍

Who We Spoke To

‍

What We Heard

Most AI features are still built top-down, and it shows

A consistent pattern in the customer-facing work is that the impulse came from above the product team rather than from a validated user problem. Leaders described executives, sales, and marketing pushing for "something with AI" because competitors had shipped it, or because an AI line item helped close enterprise deals. The product organization then built the lowest-effort thing, usually a chatbot, to satisfy the ask.

The result is a familiar one. The feature helps win the contract, and then the end user barely touches it. One leader at an enterprise software vendor described customers selecting them partly for a credible AI strategy, only for adoption of the actual feature to stay low. A leader at a B2C travel company put the homogenization bluntly, noting that every competitor had shipped near-identical AI features that all looked the same and moved little.

Several leaders described a shift in perspective resulting from poorly received AI features. The flat adoption brought them back to doing real product work, returning to the problem and asking whether AI was the right answer at all. As an e-commerce director observed, a chatbot has become a generic table-stakes feature. The interesting work begins once you stop treating "we have AI" as the goal.

The value that lands is the unglamorous kind

When leaders pointed to AI that had genuinely paid off, they rarely pointed at the marketing-friendly feature. They pointed at the back office.

The most clearly defined outcome came from a product lead at a hospitality platform, who had set a target of roughly ten percent in time saved across the product team and had realized about one percent so far, almost all of it from operational automation rather than any AI feature implementation, which he described as underwhelming. A senior leader at an automotive marketplace judged his internal AI work purely on efficiency. Did it remove enough manual work to let the business run with a smaller service team? A director at a customer-experience company pointed to draft call summaries that turn a cognitively taxing manual task into a quick review-and-tweak. A leader at a capital-markets SaaS estimated that around forty percent of his roadmap was AI-related, but stressed that more than half the investment still went into the core platform the AI depends on.

The takeaway is consistent: the value that lands is real but quiet, and it sits closer to operations than to the headline feature.

Trust, not capability, is the adoption barrier

If one theme runs underneath all the others, it is trust. A head of product at an applied-AI company stated it most directly, calling trust the single most important user-adoption challenge for AI in B2B software, and noting that it sharpens wherever the output drives a decision someone is accountable for. The word "decision," he observed, makes people anxious, because someone will be held to account for it.

The same concern surfaced in plainer terms everywhere else. Leaders described hallucination as a constant, the need to validate output as non-negotiable, and explainability as something they had to build in rather than bolt on. A capital-markets leader insisted that any AI suggestion be traceable to its sources so a user could go back and check. A product lead at an agency tool was actively tracking where his assistant fabricated or stalled so he could close those gaps. A leader at an insurance-claims platform described having to rally customers around safe data practices before they would trust LLM output on sensitive claims at all.

A useful nuance came from the automotive sector. The path to trusting AI is shorter in back-office operations than in customer-facing flows, because the cost of a single failure is smaller and recoverable. That helps explain why the durable value and the trust both tend to accumulate internally first.

"It's a tool, not magic"

The most quotable strand of the whole study is a kind of collective deflation of the hype. A senior leader at an automotive marketplace put it as memorably as anyone: "It's a tool, not magic. Saying we need to use AI more is like saying we need to use Excel more." His advice was to focus on the business problem and not go looking for a nail with the AI hammer. He told a story of a company that had stitched together an entire suite of AI tools to automate its software development lifecycle, at roughly the cost of five engineering teams, to get the throughput of one.

Several leaders said the same thing in different words. AI can solve a problem, but it is rarely the best way to solve it. A capital-markets leader described the large language model as "the one that pretends to be very clever, because that's what everyone can relate to," and warned against asking it to do the things it is bad at. A device-resale leader was the most uncompromising: "Who cares about AI? It's a means to an end. If it doesn't serve a purpose, no AI is needed." He compared good AI to a referee: if it draws attention to itself, it is not doing its job.

This is not anti-AI sentiment. The same people are building with it daily. It is a demand that the technology earn its place against the alternatives, the same way any other tool would.

Productivity gains are real, concentrated, and almost never measured

Personal and team productivity is where enthusiasm was highest and most consistent. The use cases clustered tightly: ideation and sparring, competitor and desk research, drafting of PRDs, tickets, and release communications, and rapid prototyping. The tools clustered too, around context-loaded assistants (custom GPTs and Claude projects holding the documents for a given mission), prototyping tools such as Figma Make and Lovable, and coding assistants such as Cursor and Claude Code for the engineers.

A common formulation was that AI does the first sixty to eighty percent, and the human keeps the critical remainder of judgment and originality. An HR-software leader described writing prompts that turn rough ideas into structured PRDs and transcribing team sessions into clean summaries, freeing people to be present in the room. A leader at an agency tool described conception and concept drafting as the single biggest lever, with AI as a constant sparring partner.

Two cautions ran alongside the enthusiasm. The first is measurement. The gains are assumed rather than tracked, the question of how to measure the impact has no good answer yet, and one leader rejected the premise of measuring "AI adoption" at all, on the grounds that AI is a means to an end and the end is what you should measure. The second is a craft warning from the applied-AI leader: because cheap, high-fidelity prototypes now cost almost nothing, the risk is jumping straight to something that looks finished and skipping the design process. The illusion of done is not the same as done.

The human core holds: judgment, facilitation, accountability

Asked where the role goes once the hype fades, leaders described a remarkably consistent dividing line. Execution is the part that shifts to AI. Judgment, strategy, facilitation, customer empathy, and accountability are the parts that do not.

A VP of product at an open-source infrastructure company framed the PM as the last role standing, a facilitator who spends most of his time in meetings understanding and reconciling what others need, and compared the role to that of a nurse. The job exists for the human connection, which is exactly the thing you cannot automate. A director at an e-commerce company made the accountability point directly. You do not want to hand decision ownership to an AI, because someone has to be accountable for the decision. A device-resale leader called product management "the glue between business and technical implementation," and believes that is very hard to replace.

What changes, in this view, is not whether the work exists but how it is done. The same infrastructure leader expects PMs to become communication centers, more in demand and more dependent on being organized and structured. A customer-experience director offered the sharpest near-term prediction. As developers get faster, product becomes the bottleneck, so the role increasingly turns on how quickly you can de-risk and validate while keeping the quality bar high.

The ladder is being pulled up behind the juniors

The most pointed worry in the entire study was not about senior product managers losing their jobs. It was about juniors never getting to grow into them.

A leader at an online classifieds business gave a great example. A usability study that once required three people - a moderator, an observer, and someone to handle transcription and synthesis - he now runs alone with a single tool that does it all. He called it brilliant. But the junior who used to sit in those sessions, learning the craft from two senior PMs, no longer has the reason to be there. "We are pulling up the ladder on the next generation," he said, and he meant it as a problem, not a boast.

The same concern came from elsewhere. A head of product at an automotive marketplace worried that junior PMs were outsourcing the very thinking that strategic work depends on, accepting an AI synthesis as the whole picture rather than pushing on why or what else. An e-commerce director expected fewer junior engineers and designers and a correspondingly higher need for senior people working with smaller teams and agents. Taken together, the result is a squeeze on entry-level and execution work alongside an increased need for senior judgment, which is good news for established PMs, and an open question for everyone trying to become one.

‍

Where Leaders Disagree

The most honest finding is that the leaders do not all agree, and the disagreements are more revealing than the consensus.

The first is about the role itself. One camp holds that the product manager is the durable glue and that the role will expand: more central, more strategic, more in demand as a coordinator of tools and people. The other expects the title to thin out and be absorbed into adjacent disciplines. A leader at an insurance-claims platform predicts that AI will let other roles become the PM. An engineer turned PM or a designer turned PM, rather than sustaining a separate person whose job is managing tickets and writing long specifications. A classified ads leader, who had automated much of his own product workflow, expected the role to survive but in a heavily stripped-down form. Both camps agree the judgment survives. They disagree on whether it stays packaged as a job called "product manager."

The second is the divide between tool and mindset. Some leaders see almost pure upside, treating AI as an enabler and reporting few real downsides. Others are wary of what it does to the craft. A hospitality leader states, "Toolset, check. Mindset, I'm not a fan." The worry there is not that the tool fails, but that over-reliance erodes the thinking it is supposed to support.

The third is FOMO. Almost everyone admitted to feeling it, and almost everyone had talked themselves out of taking it too seriously. The most common reframing was that much of what looks new is a repackaging of existing capability, so a solid grasp of the fundamentals protects you from overcommitting. The practical answer, in several teams, was to create ways to discuss and dismiss AI noise. A simple solution is to have a shared channel where AI-oriented ideas are shared, discussed, and dismantled when necessary.

‍

The European Cut

The study began from a hunch that the European picture differs from the American one. It does, though not in every way.

Where Europe clearly differs is on data, regulation, and temperament. Data residency and policy were major concerns, not afterthoughts. An HR-software leader summarized a constraint, stating that anything hosted in the United States tends to draw a no, and legal sits in the top three considerations on any AI decision. An enterprise software leader described the same wall on the customer side in regulated markets, where works councils and security reviews slow adoption to a crawl, and recounted solving it by hosting models in the company's own cloud to keep data inside the perimeter. Regulated and traditional sectors, banking and pharmaceuticals among them, were repeatedly described as the slowest to move.

The temperament is different too. The European leaders we spoke to were, as a group, very level-headed about AI relative to the public discourse. A travel leader floated the possibility that this is partly a bubble that bursts, after which AI settles into ordinary use the way apps did, making it a useful tool, but not the end-all, be-all.

Where the European story is weaker is on adoption behavior itself. A capital-markets leader said he saw no regional difference in how clients adopt AI, only segment differences, with banks more conservative than asset managers and hedge funds. So the regional cut looks real for the character of the role and for the regulatory environment, and much less clear for adoption patterns, which track sector and company type more than geography. There is one structural difference worth naming. A VP of product at an infrastructure company argued that the European product manager skews technical, closer to a mini-CEO of a domain, where the American equivalent leans more toward a business owner. That is a difference in the shape of the role, not in the speed of adoption.

For the more technical B2B players, one final practical pattern emerged underneath all of this. An infrastructure-and-data-readiness gate. Useful AI depends on clean data, a real pipeline, model management, and a provider layer that is itself unstable. As one leader noted, the tooling only became genuinely workable in the last two or three years, and shifted again recently toward more packaged offerings, while everyone competes for the same compute.

‍

What This Means for Product Leaders

A few practical conclusions follow from all of this.

Start from the problem, not the technology. The leaders getting value are the ones who make AI earn its place against simpler alternatives, rather than shipping it because it is AI.

Expect the durable value in the back office first, and instrument it. The gains are real but quiet, and right now they are mostly felt rather than measured. Make them visible.

Treat trust as a design requirement, not a disclaimer. Explainability, traceable sources, sensible human fallbacks, and legible states are what move adoption, especially anywhere a decision carries accountability.

Use AI hardest in discovery, drafting, and prototyping, but keep the design process intact. Cheap fidelity is not validation, and a prototype that looks finished is not finished.

Protect the apprenticeship. If seniors stop teaching through the work because AI now does that work, the pipeline of judgment that the role depends on quietly closes.

Do the groundwork that Europe specifically demands. Data residency, governance, and legal review are gating factors here, not optional polish, and the teams that get ahead of them move faster later.

Hire for judgment and curiosity, not AI credentials. Most leaders treat fluency as a baseline rather than a specialty, and more than one regards loudly advertised "AI expertise" as a small red flag. The enduring requirement is still a good product manager.

The picture that emerges is neither the apocalypse nor a revolution. It is a profession absorbing a genuinely useful tool with clear eyes: enthusiastic about the leverage, skeptical of the noise, protective of the judgment that makes the work worth doing, and quietly getting on with it.

‍