BP&R Editor Giulia Daniele sits down with Jean-Pierre Roux, Dessia CEO, about the company's innovative AI solutions, the recent three-year agreement with the Naval Group and future plans.
Dessia
[GD] How do Dessia’s AI solutions work, and what sets them apart from others?
[JR] We’re bringing to the market a new solution. Our name itself blends “design” and “AI”— in French, “IA” means AI. That fusion reflects our mission: building AI-native platforms that empower design engineers by capturing structured and unstructured knowledge.
Structured knowledge refers to the rules that are written and maintained by experts, such as how to design a specific component or system. Unstructured knowledge is the data that has been created over time and the software platform that captures it. We have libraries that gather this knowledge, and we utilise various intelligent algorithms depending on the specific case. Our platform applies data science to these components. We deliver smart automation, but that’s just the first step. We also do augmented intelligence or augmented engineering.
[GD] Can you elaborate on this?
[JR] We see two major streams of how AI is used in engineering domains. One is verification, which ensures that what you designed fulfils the task and complies with defined rules. It’s a crucial part of the engineering process. Most of the time, verification is done manually, either through 3D assemblies or technical drawings. Manual work takes longer and is more prone to errors. Our technology can automate the whole process. We can “read” a 3D model, compare it against encoded rules and verify compliance in real-time. It speeds up processes, ensures consistency and allows for continuous validation.
The other stream is generation. We use AI to generate concept architecture from existing requirements without starting from scratch. Our system can surface similar components from historical designs using signature-matching algorithms, so designers can access the database from the get-go. This is especially important if you think senior engineers are retiring. This knowledge is stored in people’s heads and disappears when they leave. Our system help capture and retain that expertise – it’s all about empowering them, not replacing them.
[GD] What would you say to those who claim AI will take away our jobs?
[JR] Like with any major leap in technology, fear is normal. I don’t believe AI to be a replacement – it’s an augmentation. An augmented engineer will be more efficient than a non-augmented one. Scepticism is allowed, but it’s not the right strategy for the future. Behind any change, there’s more than simply developing the technology; it’s a big commitment for the leadership team in terms of preparing and explaining what that change will be like.
It's not only about deploying; it’s about educating and showing that there’s value in that. Agility is essential. If you’re agile, you can achieve faster loops and be more reactive towards any project change.
[GD] I know Dessia and Naval Group have signed a three-year-long agreement. Could you take me through how this all started?
[JR] It started with them finding an article about us working on a project with AI and routing for an automotive client. They wanted to know whether they could apply the same technology to their product: ships. We ran a pilot project, which is always crucial at the beginning stage, to see if it can be deployed further.
Their products are complex but low in volume. This makes engineering costs a significant portion of the total cost, unlike automotive, where costs are diluted across mass production. They were looking for agility to optimise the development process. They used our algorithm on top of their existing knowledge to achieve faster loops and make more informed decisions.
After the pilot project, we started discussing a more global rollout across different products. This is when we signed a three-year framework agreement, which will run until 2027.
[GD] Beyond naval and automotive, what other sectors are you working in?
[JR] The solution is agnostic, but we focus on suppliers producing complex single parts. Our technology was first tested on an automotive client, but we then moved to rail, naval (both warships and civil engineering), energy (plant infrastructure) and aerospace.
Our platform is horizontal in its domain. However, as a company, we need to have a strategy. We can’t serve all industries, so we’ve identified those where our technology seems to work best, like routing. We always want to challenge the client about the business case, asking them: what is the value for you if you were to use our technology? Then, they can take us in different directions, which is essentially the platform’s DNA. We have libraries as a foundation for any AI app.
For example, we’ve conducted a digital twin project with a major OEM. They’re scanning old manufacturing sites with LiDAR, generating point clouds and using AI to recognise and reconstruct 3D layouts automatically. That’s months of manual work, done faster with AI.
[GD] What have you found to be the biggest challenge across these collaborations?
[JR] The biggest challenge doesn’t lie in the technology itself, but in the decision-making process and budget dynamics. People think AI comes with a simple click, but that’s not the case.
As with any technology, you need to educate and show the value, which is why we always conduct pilot projects with clients. Some may struggle to go past the pilot phase due to budgeting uncertainty or internal resistance. So, we try to work with clients to define their business case, so they can make the most out of the technology.
[GD] What’s next for Dessia?
[JR] From a business standpoint, we need to show our investors that we’re ready to bring this technology outside of France. We’re looking to acquire more clients. We’ve just signed an agreement with a global engineering firm, which will help us scale our technology.
From a technology standpoint, we’re incorporating large language models (LLMs) into our platform. It’s not about building our own, but adapting existing models to engineering-specific tasks. That’s the next leap: combining domain-specific AI with flexible interfaces that empower both engineers and developers.
I’d like to conclude by saying that people need to make decisions. The hype about AI has boosted the overall AI market – it’s not the time to think; it’s the time to make it happen.