Remove the Product Manager, Keep the Product Thinking: How Conveo Runs AI-Native Engineering

Conveo has no dedicated product managers. Its engineers sit with customers, define the problem and ship the feature, often with AI agents that keep working through the night. On Conveo's podcast, recorded live at Cannes Lions 2026, CTO Maarten Claes and CEO Dieter De Mesmaeker explain why AI-native engineering without product managers works for them: the product manager role went away, and the product management function stayed.

De Mesmaeker hosts and Claes, who leads the engineering team, does most of the explaining. They cover how they hire, what changed during the last year of AI adoption, why the job now looks more like steering a fleet of agents than writing code, and how the same engineering habits are spreading into customer success, recruiting and go-to-market. For insights and marketing leaders, the lesson travels well. The shorter the distance between the people who build and the people who use, the less work gets thrown away.

Engineers who talk to customers waste less work

Both founders built software at other companies before Conveo, and De Mesmaeker remembers the cost of guessing.

"Sometimes we spend a lot of time building things that customers really weren't waiting for, and they weren't really excited about."

Dieter De Mesmaeker, CEO, Conveo

At Conveo the team let customers inform the work from the start: ask what they want to achieve, work out how to help them succeed, and leave the ego out of it. Claes describes the early days as a loop that fed itself. Conveo was helping customers uncover their jobs to be done, and the team did the same with those customers to decide what to build. In his words, "we're helping our customers uncover for example jobs to be done while we're doing the same thing."

That loop shapes hiring. Claes says he looks at "engineering as product engineering" and goes deep in interviews on how candidates have handled customer interactions. Many have never worked without a product manager, so he probes for something more basic: curiosity about the problem behind a proposed solution, and the communication skills to work it out with a customer.

Asked for the three things that make Conveo's engineering team distinctive, Claes starts here.

"We have product management, but we don't have PMs. Engineers sit with customers."

Maarten Claes, CTO, Conveo

He describes sitting with a customer on a Tuesday evening and shipping a feature for them on Wednesday, because the fix was obvious once the engineer had context from many other customers and from Conveo's research impact leads. The payoff he cares about is that the feature gets used the moment it lands. The founders also describe selling and co-creating features with customers before launch, so a new feature often arrives with several companies already using it.

Dropping the PM role only works if someone still owns discovery

Claes is careful about the limits of the model.

"Just because you don't have a product manager, it doesn't mean that function isn't necessary."

Maarten Claes, CTO, Conveo

The danger, he says, is that discovery quietly erodes when no job title owns it. At Conveo, engineering leads guard the process. Research impact leads who sit on many projects push for it too. Every project ends with a retrospective on how discovery could have been better and how the team could have validated faster. "It's so easy to lose," Claes says, and he expects to keep watching it as the company scales.

This is the part most teams miss when they copy the idea. Cutting a role saves headcount. Cutting the function means engineers build quickly in the wrong direction, which is the exact waste De Mesmaeker set out to avoid.

Working with agents makes problem definition the core engineering skill

The last year changed the work faster than either founder expected. De Mesmaeker says the team was already using AI coding tools a year before the recording, and that adoption sped up sharply in the most recent three months.

Claes describes a side project of his own, a CTO cockpit that shows what everyone is working on, how many pull requests they ship and how the feature inventory is changing. "I kicked it off one day with a vision document and now it's just looping and improving itself," he says. Now and then he updates the vision document to keep it pointed the right way. The team applies the same idea in the product where it can, for example with bug reports that get fixed automatically when possible.

One engineer, Vincent, built what the team calls an agent mesh, and other engineers have since adopted it. You give it a plan, Claes says, "and then overnight it just keeps working." The engineer works through the day, prepares tickets, hands them to a swarm of agents at night and reviews the output the next morning.

That changes what good engineering looks like. Ten years ago, De Mesmaeker says, hiring tested how quickly someone could understand a code base and change it. Now the question is whether they can manage a fleet of agents and think from a product perspective. Claes, who has worked in engineering management, sees a straight line from one to the other.

"The key to being able to work well with agents is to solve the right problem, I think."

Maarten Claes, CTO, Conveo

He compares improving an agent's guidance after a weak result to an engineering manager coaching a direct report. He also notes one upside: "You don't have to have one-on-ones with your agents."

Seniority cuts both ways. Experienced engineers bring hard-earned habits, and those habits can get in the way when AI practice changes every month. Claes's answer is a growth mindset and a weekly meeting where engineers share the tools and approaches they tried. "You need to be relentlessly experimenting and some experiments will fail," he says.

The same engineering habits now run go-to-market and customer success

De Mesmaeker pushed for GTM engineering early, when Conveo had 15 to 20 people.

"The best way to scale things is to have as much of the information centralized so you can actually see what's going on in your company."

Dieter De Mesmaeker, CEO, Conveo

Most companies spread that information across a CRM, marketing automation and customer success tools, then stitch it together in a data warehouse. Conveo built one internal GTM app that covers the funnel from a first touch at an event or social campaign through to adoption. Every customer conversation is saved. When a customer mentions a feature idea and the feature ships months later, the system detects the release, so customer success can go back to that person with the update and the changelog.

De Mesmaeker sees a bigger prize in building internal tools everywhere: "building your own tools, I think it's going to be such a massive competitive advantage across the full company." Conveo now has an engineer embedded in the people team to systematize how it finds, contacts and screens candidates. In customer success, he says, internal tooling built with engineers means "one customer success person can handle way more companies than what it used to."

What insights and marketing leaders can take from a team without PMs

The engineering story maps closely onto insights work. Four points stand out.

• Shorten the distance to the customer. Claes's engineers ship faster because they hear the problem firsthand. Insights teams that put marketers and product people in front of real consumers, or in front of raw video and verbatims, cut out the relay between the person who heard the need and the person who acts on it.

• Name an owner for discovery. If your organization drops a research step to move faster, decide who still asks what problem is being solved. Conveo's answer is engineering leads, research impact leads and retrospectives. Yours can differ, but it should exist.

• Close the loop with the people who asked. The app that tells a customer their requested feature shipped is the operational version of a habit good insights teams already have: showing stakeholders what changed because of what customers said.

• Treat AI as a multiplier for judgment. Agents raise output only when someone has defined the right problem. The same holds for AI in research. Faster synthesis is worth little if the brief asks the wrong question.

For a deeper look at discovery as a continuous practice, see our guide to generative research. For turning customer conversations into an ongoing program, see how to build a voice of the customer program. And for the go-to-market side of the same shift, read Tom Husson on building go-to-market as an engine run by agents.

Customer closeness is what makes speed worth having

Conveo's engineering model rests on one bet: when the people who build sit with the people who use, speed stops producing waste. Agents multiply output, and a team without product managers can ship a feature the day after hearing the need. Neither works unless someone defines the right problem, which is why Claes insists the product function stays even after the title goes.

The same principle sits at the center of what Conveo offers its customers. Conveo's AI research assistant runs video conversations with real consumers, so insights teams can get to the why behind behavior faster. Inside the company and out, the order is the same: hear the customer directly, then build.