beti: The French AV Operator Behind WeRide's First European Robobus | Benjamin Beaudet, CEO beti

(upbeat music) - Welcome to Autonomy Insiders. The show where global industry leaders in autonomous driving unpack their real world insights. - We add two new technology to our portfolio. And what we are seeing is that the gap within the leading practice become quite narrow, okay? Once you go beyond eight hours to for operation per day, and especially when you move from five days a week to seven days a week, the economies can improve very quickly. I'd think that with this kind of condition, building the operating cost can be around 30% lower than a conventional service with the driver. So the real value is not simply replacing the driver, but it's the ability to provide more hours of service all days per week with a different creating infrastructure. (upbeat music) - Welcome back to Autonomy Insiders. I'm your host Daniel and most conversation about autonomous vehicles are conversations about vehicles or software or even sensors. But somebody also has to answer the phone when a passenger stuck or somebody has to talk to the fire brigade. Somebody has to hit the KPI is written into a contract with a transport authority. And that is a service layer and it's unclamorous. It's deeply local and it may turn out to be where a lot of the value in this industry actually sits. And my guest today runs a company built entirely on that bet. Benjamin Bude, CEO of Beti, a French operator of automated mobility networks. And Beti is a spin-off from Bertholame, a family coach business that has been moving people around rural France from more than 70 years. So Benjamin, welcome to Autonomy Insiders. - Thank you and I love it, bye-bye. - I want to start somewhere slightly unusual because I think the origin of Beti explains also maybe the strategy behind. So before we go into the history and strategy, can you give me a plain version? So if I'm the head of a mid-sized transport operator and I have never heard of Beti. So what do you do and what do I buy from you? - Quick answer, we provide the digital and human services and needed to operate automated mobility transport system for passengers and goods across Europe. And Beti did not start as a technology company, as I said. So it started inside the Autonomy, a family coach operate that has been running buses in rural France since the 1950s. So what did your family see in 2019 that made you think the right response to autonomous vehicles was to start a new company rather than wait and buy the technology later? - If we back in 2019, very few players were looking at autonomous mobility in rural areas. And at the same time, we could already see that digitalization alongside the energy transition is very important for our traditional business. So given our background and our DNA as a family company and transport operator, we feel very well positioned to explore this new segment of automated mobility in rural area. And several years later, a tool thing that there is now a broad consensus that autonomous mobility will be useful in this kind of area. So it was an intuition at the beginning, a part of our story and now a great business for us. - And your own positioning says or something like, driving is digitalized but mobility is still operated. So can you work us through the moment that became your strategy rather than just a phrase? So was there a specific incident on a route where you realized a hard problem was maybe not only the driving? - It wasn't something like a out-problem or incident. It was based on an analysis that we have done in the company. We carry out a lot of public transportation and we raised that a very simple question. If the driver is not longer on board, we are from a hundred of tasks that it doesn't relate to driving a situation. We make some study and identify 421 tasks which are made with our conductors every day but are not driving tasks. So when we discuss these issues with autonomous driving technology providers, it became clear that quite rightly, the focus was on the driving technology and only driving technology and not operating this kind of services. That's really where our strategy came from autonomous driving technology can replace the driving test but it does not replace the operator. So we need to develop digital tasks and competencies around the services for the autonomous mobility. - And you also run BT alongside your existing poach business with your brother. So how much of the autonomous existing operational muscles like drivers, depots, dispatch customer service actually transfers to an autonomous service and how much of it turned out to be maybe also not transferring. - For us, it's a real source of fragility. For BT, Bertrami is a real source of fragility. For an operation like Roland Garros, for example, we can mobilize around 10 experiences transport drivers for more than a month. So it's easy to find people from around the year to be ready for this event. And we think that there is more than 20 people, actually, expert of operation on ground situation. It's very interesting for BT to take these resources in Bertrami and use them for our operation or for our customers. - That's my accent. Maybe now let's get concrete about the product. So we call it a turnkey solution. So maybe give us the contents then. So what physically arrives when an operator signs with you and which of those pieces do you build yourself maybe also and versus you source it from somewhere else? - Our thing is really build around two things. People are on software. First, we provide teams with expertise in deploying and operating autonomous driving technologies. Such we arrived technologies. They bring the operational know how needed to turn this technology into an actual transport public or private transport. Second, we provide software licenses from two of our platforms. The first is operator vision, application suite design for the operating team. The second one is hyper vision, which enables this autonomous transport system to operate and monitor it remotely. First software are developed entirely in-house. This is also personal reason because behind before to join my browser in 2018, I spent more than 20 years building and running digital development companies. So from the beginning, I strongly believe that this software layer was strategic for the company for Betty and that we needed to own our development. So we have field team, human competencies, but also software that we develop. - And you use the word hyper vision rather than supervision or tail operation, which I think are terms that others use or so and your definition is that hyper vision cannot drive the vehicle but can react to alerts, stop it or monitor the service. So why does this distinction matter so much to you? And is it a safety argument, regulatory argument and economic one? - I think that for us, it's really about making the responsibilities very clear. With our solution, we are not responsible for the driving itself. That responsibility remains to the autonomous driving technology. Our responsibility is to operate the transport service that the technology is supposed to deliver. That distinction is fundamental for us because driving the technology and operating a mobility service at two different jobs, requiring different technologies and different expertise. And I think that global leaders in autonomous driving technology, whether they are American or Chinese, inclusive, recognize that. And they also recognize that in Europe, we have very strong expertise in operating complex public transportation service. We have big company around that. So it's natural for us to propose these kinds of services. - Maybe one more question to this hyper vision. So is this then always located within France as well, the hyper vision? - We have some regulatory to ask us to be in the same country. So actually, we have some collaboration in France, Belgium and Switzerland. And when we are in Belgium, the hyper vision is in Belgium. And when it's in Switzerland, the hyper vision is in Switzerland. - Makes sense. - Your pitch to operators is also that you train them to run the network themselves. So that sounds also like you're building a business where the customer eventually does not need you that much. So is that really the model law? Is there also a law that you always keep? - I think that the software law is really the key for us. Our goal is to train the teams that will operate the service on a daily basis and make them autonomous very quickly, quickly as possible. We keep an eye level of expertise to support them when they need it. But we don't want them to depend on us for a day-to-day operation. What remains over time is the software layer, of your vision and hyper vision. and they are the common trees that connect us to the operator, to the life of the service. So yes, we deliberately helped our customers to become operational and autonomous, but the long-term relationship is built around our software and our expertise. There's a lot of tasks that you have to do when you operate such a service like present-to-management event handling, select commitments with cities, first responders, procedures, and stuff like that. So which of these tasks do customers underestimate the most when they first come to you? Actually, because we work with a lot of mobility operator, I will say that they didn't underestimate the issues. In many cases, the fact that they understand that complexity is precisely why they come to us, and why they ask to be treated to work with them. We are still seeing a learning curve around the autonomous driving technology side, but they begin also to understand that they have to work with us, they have to create some API, some solutions to build together the solution for the final customer. So I think that actually it's more difficult with technology provider than final customers. And if we would zoom out to look more on the structure of the industry, so can you break the AV mobility business into different layers as you see it, and then also where do you see Betis sit, and maybe also where does Betis explicitly refuse to go? It's not a easy question, Daniel, because you can divide the autonomous mobility value chain in very many different layers. Since the last five years, I've seen a lot of models. If I would like to simplify in four main ones, the autonomous driving technology, the physical vehicle, the physical infrastructure required to operate the service, and finally, the digital infrastructure required to operate it. And Betis, we have made a very good choice. We didn't develop any driving technology, we don't build vehicles, so we work on our infrastructure with people and with software. And I think one number, everyone in this fleet operation layer wants to know and nobody publishes us how many vehicles can one supervisor, or maybe in your framing, the hypervisor, handle today in your operations. And what do you need to be true to get that ratio also to where you need it to be, to be economically, for example? I will ensure at the beginning with a little twist to say one vehicle at one time. But beyond that, we have actually observe two different approach. The first one, which is closer than our model, is to seeing around four vehicles to one supervisor, when a supervisor. With overflow procedure, in place of something happens on a song run vehicle at the same time. The other approach is more dynamic. Vehicles are assigned to a browser in a real time, depending on their statue and whether they will cry, human orientation. So it's very interesting to see the boss. At this stage, I don't think anyone can say, which certainly which model is the best, the most efficient at scale. As fits grow from a few vehicles, it's easy to have four to one. But when we will have more than hundreds of vehicles, I think it could be a good question. So I think that we will have the answer in the next few years, maybe in 2028, to choose the best model and the best vertex around that. And what is the actual constraint on that ratio right now? So is it the vehicle or the tooling regulation or also the human attention span? Actually, we follow a human intention and to be sure that they can manage four vehicles at the same time. We have made big, big works with some French university around this asset, to be sure that they could follow that. And we change the team after two hours of duty tour. I think it's a good one. And they have a break. And after that, they go back to the repair vision center beyond the hypervision team. So what does the crown crew look like? So we have also other tasks like cleaning, charging, depot, brutal recovery or something like that. So what share of your real operating costs is the part no party puts into a press release? We are actually quite close to the ratio that we already know from our traditional transport industry, around 20% I think, for the operating effort dedicated to this kind of activities that you describe. Autonomous mobility does not make sense without this kind of issue. So we need to develop and to budget that. And if we now look at some of your partners, for example, we write or renews. So what stops those partners from deciding maybe in two years that the service layer is super valuable and they build it themselves. What makes it bait is so hard to replace for them? Good question. What we are building is based on 75 years of real world experiences on the road. So operating transport services, dealing with procedures and working with public authorities and political decision makers, it's not easy. And I don't think they are ready to take this job. So maybe one part of the answer is that it's really not the most shining job to do. And maybe not the way that to create a lot of values that they need for their finance management or something like that. Makes sense. So here your family own structure basically also helps you in that kind of regard. Of course, of course. Maybe let's get into this deployment or into your deployments that put you also on the map in France. So in the last February at 2025. So we write first fully travelist commercial robot service in Europe. So how did that partnership actually come together? In October 2024, we traveled to Ganzhou and met with we write management team. We command them that they could be the right partner to support the entry in Europe market. Not too big, not too small with a lot of experiences. That experiences we had already built autonomous mobility and combined with the expertise of our team with an international environment time because a lot of part of my team speak better English than me. So it will be easier to build something with them. And I think that's the combination between expertise and functional experience with the key factor to moving quickly and making this project a success for the for the European market and for real. And the route you take links TGV station, a long-state parking and a business park with 150 companies. So that is a very specific commercial choice compared to your rule route, so to speak. So why that route and what did the ridership data tell you that surprised you maybe? Actually not really because this route is only about 20 minutes from our headquarters. So it's not a railway station in the big city. It's a real area. So the choice was mainly about finding the right choose case. One that we arrived was comfortable with. That local authorities can support. And for our teams that we call what that we could operate with confidence. That's why we are making this choice also because there was no existing public transport service in this area near a balance TGV which also made it easier for people to understand the value of the new service. Adoption I think it's very important in this kind of a project. So it was a very very very positive to to create something for the for the users. And what was particular and interesting was how quickly some passengers became ambassadors for the service. They came very quickly explained the function, the park, the park, all things. For me it's public acceptance is very important. So if you find very quickly some people to help you to to defer the service, you win the business. Absolutely. So they had more or less the same experience than like people in Waymo. So I often read in reports that if you want to drive a Waymo then you are like this is amazing. I never want to drive another vehicle like and it's like the same before your bus road. Yes, but with our with our influence. So we need real people that use the service every day find useful service. I think it's better not for the image. And if we look at the permits I read that the travel permits also came through in about three months or something like that as a record in front. So what did you do differently in that application process that it was so fast? By that point to be clear we we had already four positions for all the autonomous vehicle operations. This experience made a real difference. The quality of the application that we have made because we know the question that they ask us. We prepare the good applications. So it's easier to to discuss and to obtain the authorization if you are ready and you have really experimented of this kind of part of the process. So for us it was not a shortcut. It was really the result of several years of experience and the accumulated experience and it's recognized now in France but also in other countries in Europe. So we are very happy to manage this part of the of the business to be sure that we have the good authorization to be on the road. Super interesting. and we mentioned that you integrated re-write and re-write as a Chinese driving stack, and integrating a Chinese driving stack into a French public transport operation. So what generally has to be rebuilt? So was there different requirements in terms of data, flows, incident reporting, or stuff like that? So where was maybe a friction in regards to that? Absolutely. We were starting with a system that had originally been designed with many components and processes specific to Chinese markets. Because the Robobus was only in Chinese markets, they use robot-axe in Middle East, but it was the first time for the Robobus for the European market and outside Asian market because they developed some process in Singapore or the area. So we had to work on many particular issues from long-range, antifas and regulated data protocols because we have some differences between France and China. And to be transparent, everything was not perfect at the beginning of the operation. I think that it takes one year to be sure at the end that we have closed all gaps and adapt the system and making it very robust for the European market. I think that after Robaltan Valence TGV, we have made four new operations with RIDE in Europe and we have very fast for this new operation because we have accumulated a lot of experiences for the first deployment. And I think that the solution is a both solution between Betia and RIDE is very clear for the European market. So it's super interesting. Can you also give us some more context in which other places you operate with RIDE? Like the four, you mentioned. We have made Arleon, we have made Roland Garros, we have made Chateau, and we prepare some new developments that won't be announced at the time. And you mentioned earlier that you have had experience with others operators. Before you have operated, I think Navia, also Easy Mile, and then we write vehicles, so if you compare them as an operator, which one was the easiest to run? It's not really fair for this comparison because this technology I think they belong to different generations. RIDE is clearly the easiest for us to operate because it's the last generation. Avia and Easy Mile were the first and having on this sector. It's a little bit difficult to evaluate from the same way, but actually we process with our processing, we add two new technologies to our portfolio. And what we are seeing is that the gap within the leading pack is becoming quite narrow, okay? The RIDE are really the leader, and to evaluate this technology from an operator perspective, we have developed our own framework based on different criteria, more than 20, and after we have a lot of questions, a lot of tests to assess the technology. But we are moving away from simply question, which technology do I have best? To real question for us is, which technology is the easiest, safest, and most efficient to operate? That's, we search actually, we continue development with RIDE, but we prepare our new collaboration with technology leaders. It's interesting. And you also mentioned already the Roland Garou deployment, and I think it was in 2026, you were the vehicle operator for that service, but the service had been running since 2024 with other arrangements. So what made them bring you in for year three? And what does this 15-day event also with this heavy crowds night operation also teach you that your year-round route does not? To be transparent, we already, we were already the operator at Roland Garou since 2025, so it was the second year for us. What it changed in 2026 is that we were responsible not only for the operation on the ground, but also for the operation. So for the first time, 100% Betty team manage the entire operation from end to end. For me, demonstrate two things. First, the level of trust that we are now plusing Betty as an operating partner. And second, it shows that we as technology is generously portable. It can be deployed in Europe and operated locally by your parent partner without going. A lot of we write teams for the two-day operation and services. So I think it's a good issue for us and also for them and also for us because we have expertise around that. So we can prepare next operation with new customer for the bus. Another question that everyone asks is so who is your customer and who writes the checks, so to speak, is then the transport authority, public transport operator, private site owner. So, or is there even a mix and has the mix changed? Our customer is clearly the organization that provides the mobility service to the end user. Depending on the project, that can be a transport authority, a big transport operator or a private organization. We have different situations. What has changed significantly is the funding model. Today, our operations are not longer dependent on subsidy programs. The customer pays us to operate the service. We do however still benefit from France 2030 funding from some software development, but not for operations. So I think it's very important to separate the bus. Our operation is increasingly driven by commercial demand. Why public innovation funding still helps us to accelerate some of technology development, for example, on CBR security because we have a sovereignty and all things around that, so we need to be accompanied to scale our solution. And if we take now a development service as an example, so if you compare maybe a small diesel or electric minibus with a human driver on the same route, so on a cost per vehicle kilometer or cost per passenger, so where is the autonomous option today and what has to change for it to win if it does not win already? I think that today, the economic advantage really comes from the ability to operate the service over much longer hours. Once you go beyond eight hours for operation per day and especially when you move from five days a week to seven days a week, the economics can improve very quickly when there is no safety operator on both. I think that with this kind of condition, believe the operating cost can be around 30% lower than a conventional service with a driver. So the real value is not simply replacing a driver, but it's the ability to provide more hours of service all day per week with a different operating cost structure. I think that could be a good model. We also talked already about other providers, for example, that you work with Navier and Navier restructured and I think several shuttle manufacturers have failed or have been absorbed. And as an operator, you have watched your suppliers, maybe also disappear. So how does that shape your contract and your fleet strategy now? We learned something from everyone of those collaborations. So I feel more sadness for the teams who worked incredibly hard on this project than regret about the shows we made before. Today, of course, a financial robustness is one of the criteria we consider when we select the technology partner. I told you that actually we work with two new technology and it was an assessment that we have made. But it is not only the only one and certainly not the most important for us. Our strategy is not to bet everything on one technology provider. So it's very important for us to build obvious operational processes that can work with any autonomous driving technology, Chinese American technology, other origin. That is also part of what makes bety technology agnostic. I use the very several times this time agnostic technology provider and suppliers because I think that we will have success with different partners or not only with one and it will be very important for us also because we work on public transportation but also on robot taxi transportation for people, for goods. So we have a lot of opportunities to create and to build this legacy. And looking forward to the announcements of your two new technology partners. And if we look at the European landscape, the European operators that are really large, like the oldest trans staff, RATP and something like that. So are these your competitors, are your corporation partners or how do you place bety bety? Today they have become our customers and I believe we have both the capabilities and the sharing interest to be long-term relationship with them. For now, that is really our focus. We want to work with these major operators demonstrate that our model works is a good to support them as autonomous mobility moves and new strategy that they need to scale because they will have more opportunities than us. So rather to speculate today about who might compete with whom and other things, I think that we need to see and wait for the next. the first large scale deployment and after that I think that we will have a place on the market and our respective world, which would be much cleaner, clear. Since and maybe now let's look for the last question into the future. So in three years from now, for example, so what does a successful outcome for BT look like in numbers? For example, like in vehicles that you operate, countries that you operate in, something like that. Today our business plan is primarily based on the number of hypervisionally-sense deployed, hypervision for a vehicle. So the number I will give you for 2030 is 2000. That would be a very clear measure of success for us because for the biggest factor outside our control particularly in public transport. The development is very strong actually and we think that it will be interesting to have new players on this market. We currently have a significant lead for the robot bus. Robotaxi is another competition but for the robot bus it's very interesting but I think that one player rarely makes a market so they need to have some competitors and I hope that maybe actors like Olan that you know in Germany it would be a player in the next few years. So I think that if we have several credible vehicles and technology providers competing and deploying at scale that would accelerate the on-term market and our ability to reach this target of 2000 licenses. Yeah sounds super interesting and I hope it all develops the way that it comes true that you have 2000 of these licenses and in general autonomous mobility accelerates and as we talked before we started this podcast we are witnessing like the first accelerations today and I'm really happy to see that. And yeah so Benjamin thank you really much for taking the time today to shed some light on BT and also give us some insights in French fleet operator which I definitely will include now the insights into my fleet operations overview. So thank you really much for taking the time and yeah so happy to have you on today. Thank you for your question and thank you for addition. Au revoir.