The Future Factory: Where 1 Engineer Runs 1,000 Robots

Sean Weisbrot: What will the factories of the future actually look like and who's building them? In this episode, I sit down with Renan Devillieres, founder and CEO of OS Ventures, to unpack the massive shifts happening in manufacturing. From automation and reshoring to evolving consumption habits, we explore why the next generation of factories will rely on humans. And machines together, how software is being scaled to power, physical production, and the mindset entrepreneurs need to have in a deglobalization world. Renin also shares hard won lessons on deal making, system design, and what it takes to really move an industry forward. How realistic is it for the US to attempt to reshore manufacturing capabilities?

Renan Devillieres: It's not about being realistic. It's just so important from a national defense and sovereignty point of view. That has already started and will happen. The question is how,

Sean Weisbrot: how will it happen do you think?

Renan Devillieres: I think reassuring implies some sort of doing the same thing but in a different place. What I think is that the factories that are going to be created in the US are going to be vastly different from the factories we are seeing in China or Indonesia or India or whatever, because the set of constraints is vastly different. Cost of labor, availability from material and everything. So there will be factories in the us. They will look vastly different from the factories we are used to.

Sean Weisbrot: How are they going to look different? In my mind, I feel like they would be automated. And so even if you're bringing the capability back to produce this finalized high level quality product, there's not gonna be any people to make them. Do you agree with that?

Renan Devillieres: Partially. I don't buy the lighthouse model for factory. Like the, the joke was. Every factory would have like one guy and a dog, and the dog should bite the hands of the guy. If you touch one machine, I don't buy that in neither of those markets. What I buy is a high level of leverage for humans in manufacturing environments. Let me take an example. If you work at Facebook, you push code for the new speed. You click enter and suddenly the value you created is used by 1 billion people. If you are in a factory, you code one machine and then the machine works better, and then you just had an impact on maybe like 500 products per day, which is very small compared to Facebook. And so. We are seeing the first signs of this changing and factories acting more like tech. Uh, for example, in, in a Tesla factory, when you push code to a machine, you don't push code to one machine. You push code to all the machines from the California Gigafactory, but also the Shanghai Gigafactory, but also the German Gigafactory, but also the text and gigafactory. And so the leverage of a human is high because you're pushing code to a lot of machines. And so what we are going to see in the next, yeah, 15 years. There's a whole lot more leverage for humans, but it it'll be as, as, as if you were saying that there's no one working in tech. Like there's a zillion people working in tech. They just have a lot of leverage.

Sean Weisbrot: You've given a very specific example of Tesla, but the reality is that vast majority of companies manufacturing don't have plants all over the world like that. Except maybe if they're Foxconn, but Foxconn's job isn't to produce a single thing like Tesla's. Their job is to produce many different things for different companies. Apple being one of their clients. So how can that leverage actually be realistic?

Renan Devillieres: So this is actually a very good question. One of the things that I think Pro is going to change is the product mix. We have a lot of different products right now. Most of them suck. Like the door that's behind you probably was made in a factory and frankly, the level of innovation in door making in the last 40 years is zero, and there are a lot of different doors, and so zero product innovation and a lot of different products. A sure recipe for factories for having a lot of issues. Everything. Everything. If you look at smartphones, a lot of innovation. Very few models actually. Like everyone has a smartphone and they're like. 20 models making up 90% of the sales. And so when you have that, you can have very automated, um, factories. You can have lower level of diversification. You do the same thing with every process and everything. And then adaptation to each and every user is done by software. And so what we're going probably to see also is that reassuring, is not reassuring the same. Factories, but it's also not reassuring the same products. And so you can reassure when you rethink the product line to make it less diverse, but more captured to each other's via software.

Sean Weisbrot: Yeah, I agree. It doesn't make sense to bring all of the products back because the Dollar store items are not gonna be made by an American factory. They're just gonna cost a hundred dollars to produce, so we can't sell them for a dollar in the store. But at the same time, we have to ask ourselves, do we need dollar store items? Do we need those stores? Do we need those products? You know, I think Trump said something interesting, which I disagree with, like everything he says and does, but he said this one thing, and I was like, actually, and what he said was, why do you need 30 dolls? Why isn't five enough?

Renan Devillieres: Yeah, that's, that's actually very good. And also the world is deglobalization, like why everywhere and everything. And so. Probably also the consumption pattern of different parts of the world will differ more in the next 10 years than before. Like if you looked at Europe, Africa, China, the us, everyone was kind of behaving the same way. Like your clothes, like you take them by sheen, it's like five bucks and nobody gives with them. Um, that's. Model is probably dead. And so probably the level of consumption would be vastly different. Maybe buying less things, but higher quality and having less diversity.

Sean Weisbrot: I think what we've lived through in the past few decades is the ification, like basically American companies and American mindset has enabled globalization.

Renan Devillieres: Yeah,

Sean Weisbrot: and I. People are now questioning whether that's the right model and deglobalization is, I think COVID was really the start of, it was this idea that maybe we don't need globalization. But what I've also seen, which I don't really want to go into, but I have no choice but to say it, is this concept has also driven a lot of far right movements politically across the world, especially in Europe, which is quite scary. Although I, I'd really rather not talk about that.

Renan Devillieres: Here. I think there is a lot of people who got the. Of the last 20 years, uh, people who had factory jobs, not in a big city, not working in tech, not working service, not working in banking, like they were doing a perfectly good job with their factory and it got shut down and like their whole family had no job. And they are really pissed about what's happening. Take it like on reading the. Something in Brittany, which is a part of France and an industry. I wanted to put a new factory over there to make pork meat. It faced immense pressure for like, from like NGOs and people like that, saying, nah, we don't want that in our backyard. It's bad for the health and blah, blah. And I was like, dude, like you are using a smartphone that was made in China to tweet about this. Your smartphone was made. Using ecological and social norms that you will not accept, that you don't want for your daughter, and you literally don't want a factory in your backyard, but you use the products and all the people have a factory in their backyard. And they are making the products that you use. This cannot work. Like we need to have some level of coherence. Like what I would like, what I'm working on every day is trying to have a coherent world where I am okay to have a certain type of factory in my backyard. And I am consuming products that are made by people that I know, and we roughly have the same standard of living as me. And this actually. Is a very disruptive idea. Turns out,

Sean Weisbrot: I don't know if I should say this, but it came to mind. People say the same thing about nuclear energy. Yeah. Which is like, I don't want a nuclear power plant in my backyard, but like it's the cleanest energy and can be the cheapest energy over the long run. Especially if you get into the small modular reactors that don't take years to develop. It's

Renan Devillieres: exactly the same. There is that French philosopher. Uh, he died recently, like five years ago, and he was saying that the Western society was in a precent phase because it's the phase where you don't accept the consequences of the things you do, and you think that you can live in another world. Devoid of the consequences of what you do. And so you want cheap and clean energy. You want a certain standard of living. You need to be okay with a certain level of production. You need to be okay with a certain level of risk, which is nuclear and things like that. And so you are seeing a lot of political movements, but also of public discourse and also of just acts with people very pissed because they're getting poorer, but they're not producing the things that they use and they don't want nuclear. Were no factories in their backyards, but they want the standard of living. And I'm like, this will maybe last for 10 years via an insane amount of debts, but then it's reality is gonna hit at some point.

Sean Weisbrot: Yeah. So how can we

Renan Devillieres: do better? It's a mix of a lot of things, actually. The not so cool thing. Is that we will be forced to do better for one reason and one reason only, which is war. The US is uh, in a spot where they are add on against China. Russia is getting war in Ukraine. Everything, everything. And so historically factory and the production model is always tied to the war efforts. The only reason why there are so many chips factory. Getting back to the US is because war is chips and chips and is war like if you put chips on a drone, you basically have something that is better than airplanes, and so we will be forced to change our production model. Because of that, we will also be forced to change our production model because to compete and to not be dependent. Because though our tensions are getting getting up everywhere, we will have to change our operation. So we are past the phase of what can we do? And we are in the phase of we have to do it. And that's why the US is putting so much money, um, in reassuring, uh, hey, that work, but putting new factories to make chips finally in the us there's no reason. It's just. If they don't do that, they won't compete in the next 15 years. And so a lot of those, this is why, um, energy and nuclear reactors are popping up in the US is because of ai and it's because you have to prepare for, for that economic and full blown war. So, yeah.

Sean Weisbrot: Yeah. I saw Meta signed an agreement to restart a nuclear

Renan Devillieres: plant and Bill Gates, um, and Bill Gates is putting like fusion reactor next to data centers for Yeah, so it's, it's everywhere and it's just, it's fundamental. I, I really think that we are overstating how much political people can achieve and it's just like very fundamental. Forces that force you to do things.

Sean Weisbrot: I think the US is unique in this regard because of the private sector's ability to pay for change, where a lot of other countries require the government to have the willpower to make change happen.

Renan Devillieres: Yeah, that's true. That's why also the EU is, uh, is such in a, in a dire spot right now. Is that why

Sean Weisbrot: you choose to spend your time in the

Renan Devillieres: us? Um, no. This is the reason why I am spending two third of my time in Europe to create things and one third in the US to sell the things we created. And I'm back in Europe because I want to help the country that pay for my studies when I was young and broke. So I felt, it felt like a moral obligation.

Sean Weisbrot: There's a lot of people moving to Europe, especially from the us. I've seen it here in Portugal. There's no way getting around to that. What can someone like myself, who's trying to find my place in Europe do to help Europe beyond just living here, paying taxes and spending money that I earn outside of Europe?

Renan Devillieres: I really think that the US exceptionalism exists. And I think it's a mix of faith in technology, the future and the ability of people to do that. A healthy distrust of government when it's getting too big and willingness to take risks and be entrepreneurial. And so I really think that Europe is lacking a lot of those parts. And if you want to help Europe. I think this is what you need to bring politically in terms of what? Specifically? Entrepreneurial energy, uh, a, a healthy dose of skepticism around big governments. I constantly don't trust governments. Yeah. But I, if you go to a, if you go to a meeting of entrepreneurs in France, you'll get asked. What is your relationship with the states, which is like, why does that matter? Because the state is so big in Europe, they're the ones who decide, they're the ones who make the law. There is a lot of laws and they also control a large part of the money. Like the biggest investor in France is the French public bank, which is bonkers if you ask me. So yeah, there's, it is the rate.

Sean Weisbrot: One of my friends has a company that's, he started in China, French guy started in China. And he also has a French company, so he spends half of his time in China, half of his time in Paris, and he got his first half a million in funding from the French government and then raised money from Chinese investors using that, like saying, Hey, I've already raised money from the French government, which is why he needed the French entity, was to be able to receive the money, but he needed the Chinese company in order to receive the money from the Chinese investors.

Renan Devillieres: That's very typical.

Sean Weisbrot: And he also involved in manufacturing his business uses AI to detect fraudulent products, defective, uh, like products that are faked. So like copy. Okay, that's good. Copycat products. And I, I interviewed him years ago on the podcast and he actually helped to take down a, a ring of, from the mafia in China that was creating fake products, that medicines that were actually killing people when they took them. That's so cool.

Renan Devillieres: Yeah. Congrats to him.

Sean Weisbrot: What are some of the challenges of the work that you're doing?

Renan Devillieres: I would say my number one challenge is the human capital. So our business is that we are creating software companies for the world of manufacturing. We created 22 in the last four and a half years, and at the end of the day, it's all about having the right human capital. And France is particular because we have really good engineering talent, but. Don't have exposure to at scale, um, development. Like there are very few French engineers that have managed product with like half a million users. And so bidding for half a million users is very different from doing a cool demo to sell to someone. So that would be my change. Number one, the change number two. Would be product thinking, because I think what we're good at, but that's very hard, is taking very specific challenges of very specific manufacturers and inferring what is the generic solution to that and how tech can help. And so if you are not solving the right issues. Well, you don't have a business, but if you are solving the right issues, but you are too specific, you don't scale as a tech company. And so scalability comes from being generic enough in how we can treat the thing. And so that's a constant uphill battle. But when it works, it's really cool. One of our companies, they just signed a very large fundraising in four years, they deployed 800 factories. Their largest client is running 30 K daily. Users like interacting with them, so that's really cool when it works. That's, yeah, that's the tech challenge.

Sean Weisbrot: I wanna go back to something you had said about the engineering pro, the engineering mentality of developing for half a million users. What is the difference in the thought process and engineering needs to have? I'll give you an example why I'm asking this question. So I've, I've built a software for businesses to cut their costs, and it's a, it's a simple program, but I built it myself using AI a hundred percent. All of the code was written by an ai. Using my prompts, and so I, I understand product and design and, and all of these things, but I don't have a background in architecture. I'm, I'm, I'm not classically trained in computer science. I don't have architectural experience except for this, so I can't even tell what my software is capable of supporting. I don't know if it could support a billion people. Probably not, obviously, you know what I mean? I don't know if it could support five concurrent users or 500,000 con, so how, how can someone direct an engineer to think for scale?

Renan Devillieres: So there are, there are, there are two parts. The first is product and the second is system thinking. In terms of products, there is a lot. Of the context of each user slash each company that will be different from one to another. How can you capture the context with as few things as possible, but the user feels like. It's my context, and I agree. And then so there's a context of input and then there is output. How can we take that user and put him or her or she through the right steps to get to an output that is exactly precisely what he or she wants. But we'll also match with the context. So that's the thing that's different. So for example, let's take an example that everyone knows A CRM, like you have funnels and steps and everything. Some clients will have 17 steps, some will have three, some will have 10,000 deals, some will have five. And so how do you make it so that you cater to right? Sometimes you have to make choices. I. You say we are. We will only focus on high velocity CT deals with a lot of users and blah, blah, blah. Because our core value is the conversation between a large group of people on a few deals, or one guy is responsible for one deal. There's no conversation, but you have to go through a ZI deals, so that's product. And then you have system thinking in tech, if you design the wrong system. Then it probably will break. And so designing the right system for the AI use case is actually a lot of math, a lot of system thinking, design and understanding on what it is. Uh, so let me take an example. One of our solutions is leveraging AI to have better process engineering. So basically you have a big machine, it has like. Parameters, temperature, run rates, um, quality of vaccine. So you have 20 things. And so you're supposed to say to the end user, Hey, end user, you need to up the temperature by three degrees and it'll run better, and okay. If you do that, then the right architecture is actually multiple architectures. You have one architecture, which is for the 10 experts in the company that will do the setting and say, when this happened, you do that because there is a rule, blah. And so that is. High latency. Frankly, you don't give a damn about latency because the expert can click and enter and goes half a signal after you damn. Uh, it's a latency barrier, but then when you have the part that's the guy on the machine, you need to be low latency. You need to handle gigabytes of data per hour. And so you need an architecture that is vastly different. It probably needs also to be local and not in the cloud because you have norms and in engineering, in making planes or making pharma products you cannot afford. I. To not run the machine or the machine braking if you don't have access to internet. So like all those things that you need to think of are very important, and then you scale it with the number of users and the scale at which you operate. So there's a lot of tinkering in that, and that's actually a full-time job. Like a large thing that we don't have enough in in Europe is called reliability engineers. And those people are literally in charge of thinking of that and making thing work.

Sean Weisbrot: So for someone like myself who has a software that is just starting to be used, how can I make sure that my architecture can scale? Once I've proven with the, the few businesses that have have already used it,

Renan Devillieres: what I would say is define the moment where you will have a significantly different budget. To scale. So maybe you say, okay, at 1 million revenue I will hire a team and things will be different. Okay? So you say, okay, so I need my thing to scale to 1 million, because there is no scaling to infinity. You scale to a certain point. So, okay, so scaling to 1 million, how many users, what are the, what is the latency, what are the expected level of service? Blah, blah, blah, blah. And then we, when you have that. You can hire like an architect or system designer for like half a day and to say, okay, I'm trying to achieve that. Please tell me the optimal architecture that I. But the first part is very important because you don't design for 1 billion user day one. You basically design for your next Mason.

Sean Weisbrot: So wouldn't it make sense for me to spend half a day with someone right now and say, my goal is to make it get to a million dollars a year in revenue. How do, is it, is it capable of this right now? Or what do I have to do to make it?

Renan Devillieres: Yes, exactly. That's what I would do if I were you.

Sean Weisbrot: Okay. Yeah, I've been thinking about this, but I haven't done it. But I think it's something that I, I should do. I agree.

Renan Devillieres: In the era of vibe coding, it's getting more and more important.

Sean Weisbrot: So then the next question would be, how do I find someone that I can trust to look at my code and give me the right information?

Renan Devillieres: Rule of zombies. I always trust people who have done it three times and at the step after my step. So if you are looking for advice for the 1 million mark, look for someone who has done it three times at the 10 million mark. That's a good heuristic that I use.

Sean Weisbrot: So a CTO and architect who started from scratch and took it to 10,000,003 times? Yep.

Renan Devillieres: Yep. Okay.

Sean Weisbrot: See, it's interesting you say that because I was having an issue with my integration. There was a bug and that thing took me weeks to figure out on my own, and I was, I was ready to either purge the code and start it again. I mean, it's a, a larger software and it was a, a feature set or to hire someone, a developer, to come and look at it. And I went to Upwork, ready to hire someone. I reached out to a bunch of people and then I, one of them gave me an idea it was wrong, but then my frustration caused the AI to go, oh, I know why it's not working the way you were expecting it, because I didn't code it the way you were expecting it, even though I told it before, I was expecting it. So then we, after a few more bugs that kind of came up after the fact, I was able to fix it. I had people, I, I said, it's a QuickBooks integration, so I was specifically looking for someone who successfully debugged QuickBooks integrations because they're notoriously awful to work with. And I had like 15 people, 20 people reach out and they're like, yeah, I can do it. I was like, yeah, but have you done it? Well, I know how to work with integrations, so like, yeah, I can do it. No, no, no, no, no. Have you done QuickBooks? Yeah. No, then I can't pay you $70 an hour to, I had people say, it'll take me two hours. I had people say, it'll take 60 hours to debug. I'm like, guys, come on. Luckily, the, my frustration helped the AI to clarify my goal, and then I was able to fix it for free in like an hour, but I was messing with it for weeks looking. So yeah, I'm, I'm happy that it's done.

Renan Devillieres: It it tough. And if you were in the US still, you would've like 10 people reaching out. In one hour.

Sean Weisbrot: Uh, so I did have two people from the US and the rest were like Indian, Pakistani. Okay. Yeah. The American was like, I'll do get, I'll get it done in two hours. It'll be like 50 bucks an hour. And I was like, I, I almost hired him. 'cause he said he actually had experience with QuickBooks.

Renan Devillieres: Yeah. And my rule of thumb is either you have done it or you, you haven't.

Sean Weisbrot: Yeah,

Renan Devillieres: both are fine. But don't bullshit me. Then on certain things which are high level decision making, I always trust the guy having done it like two or three times on certain topics like go to market, top of funnel or product thinking. I like to take newcomers and people with fresh ideas. But when it comes to like technical challenges, like, like nah.

Sean Weisbrot: So what's the most important thing you've learned from your career and your life? Wow.

Renan Devillieres: My first big learning was after almost 10 years of working, I took a vacation alone because I need to think and I said I'm good at what I do. People pay me a lot of money to do it, but I'm not that happy. I. And it's very weird because I always outperform almost everyone comparable in my position. And I'm threatened being, being, with being fired like half of the time. And so that was very weird for me and it hit me that I was actually under an entrepreneur, but I had been working, operate for 10 years, but nobody in my family or direct. People that I knew were entrepreneurs. And so I, it actually took me 10 years to understand that. So that was a big learning. And I would say that the second learning is I'm an engineer at heart. Uh, so I like systems. I like understand things. I like numbers. I like designing a beautiful process that works, everything. So I solve things and I think I'm good. But in the last 10 years, I fell in love with deal making when it's done right. To me, a deal making done right is you understand your counterparts extremely well, so you craft something that will help him or her so much that incredible value will be created. And I had like that very. Engineering mindset about deal making. Say, Hey, they, they sell carpets. Those people will lie through their teeth to get to you and everything. Like I had that very negative view of deal making. Plus, I'm European, so it doesn't help. But then especially in the US, I fell in love with deal making when it's done right, because it just drives progress forward when it's done right. And so yeah, that became, I, I actually take, take a lot of pleasure and I think the companies are benefiting a lot from my deal making abilities because I like it when it's done right. And so one, one thing that's one of my mentor told me is that ideally you want to do a good product and you want to have so many conversations that you almost like choose your clients. And you choose your clients because you're like, I can't wait to start working with that person because it's going to be awesome. And so, yeah, that deal making and relationship with others was a big learning for me.