Is AI the secret sauce that lets the West deglobalize supply chains and bring factories back home?
In this episode of TechFirst, I talk with Federico Martelli, CEO and cofounder of Forgis, a Swiss startup building an industrial intelligence layer for factories. Forgis runs “digital engineers” — AI agents on the edge — that sit on top of legacy machinery, cut downtime by about 30%, and boost production by roughly 20%, without ripping and replacing old hardware.
We dive into how AI agents can turn brainless factory lines into adaptive, self-optimizing systems, and what that means for reshoring production to Europe and North America.
- Get deep insights on the TechFirst Substack
- Subscribe to the TechFirst YouTube channel
Check out our conversation here:
Transcript: AI agents in manufacturing
Note: this is an AI-generated transcript. It may not be 100% correct.
John Koetsier
Is AI the secret sauce that will allow the West to deglobalize supply chains and re-industrialize? Hello and welcome to Tech First. My name is John Koetsier. One of the biggest global stories of the last year has been a battle to reorient global production and trade. And while, let’s be honest, the methods have been pretty much a complete shitshow in the US—and tariffs may not be the right answer, and certainly going up and down is probably not the right on-off approach—it’s arguably important for large trading blocs to have some control over their means of production.
Could AI be the key to making it happen? A new generation of engineers is doing that right now in Europe, with a layer of AI agents to coordinate and optimize manufacturing, even with legacy machines. They’re already boosting production by 20%, dropping downtime by 30%, and they just raised a $4.5 million pre-seed round to make it even better. The company is ForgIs, and to chat we have the CEO, Federico Martelli. Welcome, Federico.
Federico Martelli
Hi, thanks for the intro. Happy to be here.
John Koetsier
Super pumped to have you. Thanks for joining. Let’s dive right in. You’ve said that intelligence is the next geopolitical frontier. What do you mean?
Federico Martelli
Yes. What I mean is that it’s difficult to compete with Asia on other fronts, right? They have scale that we don’t have in Europe because we’re fragmented. The US still doesn’t have the same scale as India or China. The US especially did a pivot. They stopped focusing on becoming the best in all technologies. They went all-in on AI, right? And you see that in Silicon Valley startups, YC startups. So what they’re doing is betting everything on AI, because whoever wins the general AGI race is going to win most of the markets, also in manufacturing.
John Koetsier
Interesting. You’re also applying that to machinery, however, because you’re actually applying AI and AI agents to existing manufacturing machinery—perhaps new machinery as well. Talk about that.
Federico Martelli
Yes. New machinery is much easier, of course, because it’s more technologically advanced. However, legacy is where you have to focus if you want to build a profitable business in manufacturing. The last 30 or 40 years was an offshoring trend to the East. Most of the production went East, and China especially is highly automated—same for Japan. They own 70% of the robotics market—Japan, for instance. So you see they understand that’s their thing, because they dominate, right? I think 40% of the production is coming from there.
What we have to do is bring it back and compete. How do you compete? You compete with intelligence, because intelligence is cheaper, more efficient, and optimized.
John Koetsier
Mm-hmm. What’s that look like on old machinery that probably doesn’t access Wi-Fi?
Federico Martelli
Exactly. You have two problems, as you said. One is an integration problem, and then there’s intelligence. And you cannot solve both with the same solution. First, you need an integration layer—something that, cross-machine, cross-brand, cross-application, is capable of integrating with all machines and extracting the data, because intelligence is data. Then you need a second layer that adds vertical intelligence—to reduce downtime or enable flexible automation. Those are the two problems we’re solving.
John Koetsier
So you’ve already achieved some pretty significant results. You’re reducing downtime 30%, boosting production 20%. There was some other stat in the prep materials as well. How’s that happening?
Federico Martelli
First of all, one of the mistakes when you think about manufacturing is that everything is cutting-edge and advanced. It’s not. It’s super behind. So even slight improvements or adding some technology makes a huge difference because there are outdated systems. That’s how we were immediately able to achieve these metrics after three to six months of starting—three months full-time. And then, the more we integrate and improve our product, the more we can achieve.
John Koetsier
So is that primarily found in efficiencies—making sure material is where it needs to be, making sure you’ve got just-in-time manufacturing so everything comes together at the end? Is it mostly internal in a single factory?
Federico Martelli
Yes. First, we have an integration layer. After we’re integrated in a factory, we start adding verticals. One vertical is fast reconfiguration—flexible automation and robotics. This is very helpful for low-volume mix, so companies with low batches of production and frequent changes.
Another one is mass production. When you produce large volumes, what we do is identify—specifically with edge AI—where the problem is, what sensor and what drive caused the downtime. Then we explain to the operator: look, you should do this, this, and this.
John Koetsier
Mm-hmm. So you kind of need a deep understanding of manufacturing, not just AI. You must have engineers who have actually worked in that scenario already.
Federico Martelli
Yes, we do. And that’s a very smart question. We have a model similar to Palantir. First we integrate. We take three to six months to be in the factory and understand everything—the process, the logic. This is super important. Ninety percent of AI companies fail because they fail to integrate. They fail to understand what the pain points are, and then they die before scaling the company.
John Koetsier
Imagine that. You’re diagnosing and then you’re prescribing. Love it.
Federico Martelli
Exactly. Then we have a recurring license model where we sell our verticals. We add verticals of intelligence, so to say. Yes, that’s the model. That’s the only thing that makes sense in manufacturing, with a huge system of machines.
John Koetsier
Yeah. So you raised $4.5 million. I saw on your LinkedIn—I think you did it in like a day or something. It was kind of insane. Crazy. I guess that speaks to a couple different things: one, you built something real and it’s not just napkin stuff; two, you’re hitting a nerve in an area where it’s important; and three, you must have a great network. I noticed you had some people from Arduino, all that stuff. Talk about that scenario.
Federico Martelli
Yes—36 hours. There were two things that led us to this round. First, we had two term sheets already on the table, and they had a deadline. Having a deadline is helpful for other VCs to jump in or not jump in. We had a deadline, and we got five term sheets—three from the US and two from Europe. Then Redalpine was the best fit, and they really understood manufacturing.
Massimo Banzi, the founder of Arduino—which was just sold to Qualcomm—jumped in at the very beginning. He was with us from the very first day. We didn’t know him. We just pinged him on LinkedIn. He reached out, he answered, and he said, “Look, let’s talk.” We spoke; it was a fit. He really liked the fact that we understand the industry. We all come from there. I did consulting—my projects were in manufacturing. Camila worked across an entire value chain and did research on that. We were speaking the same language. That’s why he was with us from the very beginning.
John Koetsier
Cool. In a former life, I had some connection with lean manufacturing—the Toyota Production System and stuff like that. Does that fit into anything that you guys do?
Federico Martelli
Yes. Lean manufacturing is what we’re doing now—optimizing it, making it very efficient. That’s what we do.
John Koetsier
Interesting. And that actually involves the people running the machines in the process. Is that part of what you do as well?
Federico Martelli
Yeah, another smart question. At the very beginning, we had a system that was completely replacing people. But first of all, you need people to check because this software runs on real hardware, so it can’t damage the surroundings. You don’t want everything fully independent with AI agents. The risk is chaos and no one knows how to fix it.
Second, internally in the company, there was a lot of pushback. So we changed that quite quickly and started integrating the operator into the system. Now we supercharge the operator and the engineer to make the work better and more efficient, and give them the tools to do it better.
John Koetsier
It’s super interesting because I’ve gone through some of those lean manufacturing sessions. It’s a week-long session. The people involved in the process are part of reengineering it. You look at upstream inputs, downstream outputs, customers of each process, and then you optimize everything—positioning in the factory, pre-positioning material, whatever the case might be. The key is that the people who have to run the process after are involved in the decision-making. They’re invested. It’s huge.
So I want to get a bigger picture for a second. You’re doing some very cool stuff. It’s going to be incredible because you’re adding a data layer—a real-time data layer—which the lean manufacturing people never had. They’ve got Kanban, they’ve got paper.
Federico Martelli
Yeah, exactly. Excel sheets.
John Koetsier
Exactly. They don’t have real-time data, so that’s very cool. And you’re adding real-time intelligence too, so if things change, you can adapt really quickly. That’s huge.
Now, bigger picture: what do you want to achieve—let’s say on a Europe scale, not just a factory scale?
Federico Martelli
The macro achievement we want is being able to bring production back from the East to the US, Europe, and Canada—domestic, where it was born. There’s a sustainability impact because you don’t have transportation around the world. You produce and you sell.
Production capabilities are a geopolitical asset—very strategic, especially given how the world is now. And second, it’s important because the US and Europe are the largest output markets at the moment. We buy and consume the most, so it’s fair that we produce and sell. This is what we want to achieve—making industrial automation and manufacturing efficient, optimized, and also very flexible. That’s important, too, so you reduce waste. You receive an order flow, the system adapts, it produces. Then you receive a different volume, and again. Everything is connected with this layer.
John Koetsier
Yeah, very cool. There are huge challenges with that, as you know. Those challenges aren’t just in a factory—wiring up a factory for intelligence and data collection and automation and agents. It’s beyond that, because you need an ecosystem for manufacturing. You’re building a lamp and you need bulbs and a controller and a little chip. You need a Wi-Fi chip or a Bluetooth chip because you’re making it smart.
All those bits and pieces add up to an ecosystem, which in China is incredibly dense. You can put all these things together: that piece is made two kilometers away, that one is half a kilometer away, this one is five, this one is right next door, and boom. You don’t have that necessarily in the US or Europe. How do you solve that problem?
Federico Martelli
Yeah, I completely agree. If you think of San Francisco, it’s the same for startups—you have an ecosystem, right? You need a Silicon Valley for manufacturers.
First of all, we had that ecosystem. Fifty years ago, Germany was one of the countries with the most manufacturing. Same in Italy. Now we lost a lot, but we had it—so it’s possible to rebuild it.
We need to focus on the reason we moved it out, which was cost efficiency. I think if we can make it more convenient to have production at home than outside, most companies would prefer to have it inside. Then once this trend starts—Europe or the US, and the US is already pushing in this direction—production starts moving back, and the ecosystem can be rebuilt. At the beginning it’s going to be an effort, but eventually, industry by industry, you rebuild that density. There is no other way.
John Koetsier
Yeah, you have to start. There has to be a beginning. You probably need to go industry by industry because it’s a different ecosystem each time. But it’s arguably a matter of national security, international security, to control and own significant parts of critical infrastructure. We saw that during COVID, where countries started hoarding things they thought they needed in a pandemic. That can happen in times of war too. We already saw things with rare earths—China put a massive export ban on it and then reversed it. This is a really, really—
Federico Martelli
—and there are industries where we are already dominating, right? Like defense. There are industries where we already have everything in-house. Defense, for instance—we didn’t offshore there. So those are starting industries. Cars—Volkswagen produces in Poland. So there are cases.
John Koetsier
Yep. Absolutely. And it’s interesting because the narrative is that everything’s made in China. It’s not really true. A lot is made in the West in a lot of different places. And if the attitude changes and the effort changes, there’s value there.
Excellent. Very cool. So you’ve got your digital engineers. They’re running in a factory. Tell us what they’re doing on a regular basis. Are they controlling machinery? Are they providing inputs for people to control machinery? Is it a mix of both?
Federico Martelli
Yes. To go a bit back, we need to focus on the components. All the hardware components have their own platform, and the incentive from the hardware producers was to keep these platforms closed. So you cannot switch between different robots, for instance.
What we had to do with our digital engineers was integrate through an API—some brands can do that, like Beko—or through their user interface. So we have digital engineers that use the platform’s user interface like a human would. That’s a workaround. Then we use the native platform, and that’s how we integrate—or the API in some cases.
That’s how integration happens. We extract data and information, and we use that for the vertical intelligence—real time. We run real time.
For critical choices, we always have the operator, as we spoke about before, because you want a human check. We’re not ready to have AI agents do everything from A to Z. That’s a recipe for disaster. We always need a human check to verify, “This makes sense.” We just get the human there faster—realizing, “No, this needs to change,” or “We need to add oil to the engine,” right?
John Koetsier
Yeah, makes sense. Cool.
Okay, you raised $4.5 million as a pre-seed round. It’s not a small round if it’s a pre-seed round. I’ve got to reorient myself now because I’m hearing about these $10 million pre-seeds and $4 million pre-seeds, and then the seed round is like $30 million, and I’m going like, whoa, something changed here.
Anyways, you raised the cash. What are you going to do with it?
Federico Martelli
Yes. We raised so early because we started really a few months ago, so it was a pre-seed. We are the team, we are the idea, and we had a bit of technology from ETH that we developed during our studies. That’s why it was a pre-seed.
Now the plan is to focus on three levels: sales, product, and tech—in that order. First you sell and you get into a factory. Then you develop the product with adjustments for the client. The tech is the same, but you need adjustments because the application is different and the machines are different. So you integrate specifically for the client. And once you run, you can add the vertical intelligence. The tech is cross-factory, but the integration needs to happen.
That’s what we’re doing. We’re focusing. We have 14,500 clients, so we’re doing pretty well. Now we just need to start some of the POCs in the new year, and with small factories we already started.
John Koetsier
Cool. It strikes me as interesting—the stack you mentioned and the order you mentioned it. In the past, people would start the opposite way. They’d start with the tech, then the product, then sales. You’re going sales first: “Hey, does somebody want this thing?” Okay, then I’ll make the product. Okay, I have some tech, I’ll build some of the tech. That makes more sense.
Federico Martelli
Yes, we learned that the hard way. It’s easy to think you should do the tech first, then the product. The tech should be there, but the product first. It’s wrong to do it the other way because there are so many different ways operators work in a factory. An operator from Volkswagen is different from one from another company. That’s why we adjusted.
John Koetsier
Appreciate your time. Thank you so much, and best of luck.
Federico Martelli
Thank you very much. It was nice to be here.