Open source, 4-day workweeks and bootstrapping… Pollen Robotics, a startup going against the grain in its sector

Growing a start-up is never an easy path, and rarely a linear one. Sometimes, you also have to think outside the box, as Pollen Robotics has done, aiming to compete with the American and Chinese giants in the humanoid robot and embodied AI markets. Matthieu Lapeyre, co-founder and CEO of Pollen Robotics, explains why at the closing ceremony of Inria’s Fête des startups 2024.

“I don’t know if I’m here to set a good or bad example – I don’t think we’ll know until much later – but the point is to show you that there are multiple paths to developing a startup,” announces Matthieu Lapeyre, co-founder of Pollen Robotics, right from the outset, a startup he founded in 2016 with Pierre Rouanet, another Inria researcher and member of the “Flowers team”, within which germinated the work, but also the philosophy that gave birth to Reachy, a humanoid robot that is today the success of the startup.

“In the Flowers team, we were working under the supervision of Pierre-Yves Oudeyer, at Inria Bordeaux, on the subject of the connection between AI and robotics, and in particular the way in which robotic agents learn in the real world via interactions with humans”, explains Matthieu Lapeyre, before describing the workings of this structure that enabled him to try his hand at intrapreneurship: “We had a fair amount of autonomy and the trust of Pierre-Yves, who let me stop my work on artificial intelligence algorithms to develop a small robot project I had at the time, Poppy, which met with a certain degree of esteem, and which enabled me to take Pierre Rouanet on board in a project that has become quite important today in the robotics community. ”

The desire to work together in open source as a corporate project

This project is called Pollen, a name that came naturally with the idea of “disseminating the work of the Flowers team, but with an organic and poetic feel that contrasts with the very synthetic and futuristic themes found in other robotics companies”. Matthieu Lapeyre’s and Pierre Rouanet’s motivations also reflect a different approach: “We launched Pollen to continue working together to manufacture and democratize open source technologies in the fields of robotics and AI, via simple, elegant products, in a company made by and for ‘makers’… You soon realize that this isn’t exactly the pitch for a box that’s going to make money!”

No revolutionary product or technology yet, no identified need… Pollen is initially groping ahead, developing an initial prototype robotic arm. “Manipulation is an important topic in robotics, so in 2017 we developed a human-scale Poppy arm. A year later, we added another arm and vision to do advanced manipulation and deep learning, which gave the basis for Reachy, our first robot, which was launched at CES in 2020 and got great press coverage.” Despite the health crisis holding back the commercialization of this first robot, the team is continuing its development, developing tele-operation in virtual reality, which has since become commonplace on the market, as well as mobility. In all, this first-generation robot has sold over 100 units. “This may not seem like much, but in the robotics market, few companies have actually deployed such a large number of robots.” In addition, Reachy 1 enabled Pollen to develop new prototypes and finish second in the 2022 Xprize competition, pocketing a check for two million dollars in the process.

The bootstrapping challenge

A welcome sum, as the road has not been an easy one: “Doing science in business is not easy. We have the impression that our vision of the future and the important subjects we work on are self-evident, when investors simply ask us: what is your market? What’s the outlook for the next 5 or 10 years? We felt we were wasting our time looking for financing, so we decided to bootstrap, i.e. to finance ourselves through the sale of our products. In the world of hardware, robotics and AI, it’s quite tortuous… I’d only recommend it if you’ve got your heart set on it!” As proof: with CES 2020 in mind, Pollen was developing an AI arcade terminal, before changing its plans two months before the event to focus on Reachy and its well-known success.

But then we had to find use cases for Reachy that would interest potential customers. “The first two years of Inria funding are very pleasant, as we can focus on developing our ideas. After that, the money problems come back regularly, and the hard part begins, with its ups and downs.” A case in point was in 2022, when the startup owed its salvation to its banker, who agreed to extend its overdraft capacity from -50k€ to -100k€, before the company won second prize in the Xprize competition. “He was the first person I called to thank him,” says Matthieu Lapeyre, for whom ”you’re not dead when you reach zero cash flow. You’re dead when you decide to stop fighting.”

Marked by this exceptional year, the Pollen Robotics team made a documentary about it: Pollen Robotics: The Making Of A Challenger, available on YouTube. As 2024 draws to a close, Pollen is delivering the first models of Reachy 2, the new version of its embodied AI robot, according to its distinctive approach: “friendly, open and collaborative, in contrast to the flood of cold, futuristic proprietary projects coming onto the market in the image of Tesla’s Optimus.”

An atypical form of governance

The company is also working hard to develop an open source robotics platform that would be an alternative to these proprietary firms, in partnership notably with Hugging Face, the French AI nugget that shares the same philosophy. “We’re very proud of the team we’ve assembled, but also of the collaborations initiated with leading research laboratories that are also our customers, with whom we work to improve our solutions. All this, while respecting our values.”

These values have enabled Pollen Robotics to bind together a team of around thirty members, all “friends, or friends of friends”, recruited by co-optation and prepared to say no to salaries two or three times higher in order to benefit from a unique working environment: “We have replicated the Flowers team model, based on trust. There are no managers and no time controls. Everyone has their own CB to make urgent purchases if need be, and there’s a single pay scale. But the best part is the adoption of the 4-day week since 2021! At the outset, we were told that it would be difficult and that we’d have to structure ourselves better with 5, then 15, the equivalent of the number of Flowers team members… Now we’re 30, and we’ll keep going until we screw up,” says Matthieu Lapeyre.

In conclusion, the CEO recalls the importance of trusting oneself and following one’s instincts to find the motivation to undertake: “I have the impression that every time we were lost, it was when we had listened to advice we didn’t want to listen to. We were no longer ourselves, no longer carrying our own ideas. Of course, we shouldn’t be extravagant and listen to no one, but on certain essential points, it’s okay not to do what everyone else is doing, in order to preserve our originality. Looking back, I realize that all our successes were achieved when we were in line with our values, and had the freedom to be so. The biggest risk is not to try what you want to try: at worst, the company stops, and what will bother me most is not working with my colleagues. But if you don’t try anything, the risk is to become a prisoner of a company that no longer inspires you, with all the hassles, and without the pleasure of entrepreneurship.”