"Turning 100 engineers and developers into a unified, productive, innovative team."

Caroline Goulard:
Hello, I’m Caroline Goulard. I created two companies in the field of data visualization. I’ve been working for more than 10 years to build bridges between humans and data, and I’m very happy to be with you today for this new episode of The Bridge, the media by Artefact that makes data and artificial intelligence understandable for everyone.
Today we are with Robin Doumerc. Robin, hello.

Robin Doumerc:
Hello.

Caroline Goulard:
Could you introduce yourself?

Robin Doumerc:
I’m Robin Doumerc. I’m the CTO of Artefact, so I handle all the engineering organization there.

Caroline Goulard:
You lead a team of more than 100 engineers and developers. How are you structured? What are your challenges?

Robin Doumerc:
My team is composed of three teams: frontend engineering, where we present results to the client; backend engineering; and data engineering, where the magic happens and where we bring all the data and data science together, and where we make AI.

Caroline Goulard:
So it’s three very separate teams, or do they have bridges between themselves?

Robin Doumerc:
It used to be very separated. Last year, we structured to create one unified team in order to break some silos and find more synergy between the teams.

Caroline Goulard:
And was it difficult? Or do you have a recipe for that?

Robin Doumerc:
There’s no one recipe for it. It was not difficult, but there are still some challenges. How do we keep the culture of each separate team still very present, and at the same time allow them to have more connection between each other and find more fluidity between them?

Caroline Goulard:
When you wake up in the morning and go to the office, what are your main challenges? What are you working on daily?

Robin Doumerc:
First of all, I work on my client projects. I’m still handling many client projects and staying hands-on. They face many challenges—organizational, technical—and my first job is to help my clients.
Then I have internal challenges. I strive to have a high-performing team. They have their own set of challenges, and they come to me to help them or coach them in order to solve them and make them overcome their challenges.

Caroline Goulard:
What about generative AI? Related to the performance you’re looking for, does it change something for you? Is it helpful?

Robin Doumerc:
Generative AI is very helpful, especially when we’re tackling new challenges and new problematics we’ve never experienced before.
It used to be a bit of a rough patch when we had to learn and discover new topics. Sometimes it took us time to actually understand the whole problem very well. Generative AI helps us go through a first impression of the problem very quickly and guides us in finding steps to solve these problems.

Caroline Goulard:
So does it mean that you’re doing the same kind of work in less time? Is it productivity, or is it quality that’s different?

Robin Doumerc:
It’s a bit of both. We see, sometimes on certain types of problems where we’re discovering something new, we tackle the challenge quicker.
But also on topics that we know, we’re often under the pressure of time—we need to deliver projects quickly. Usually, documentation or best practices can be a bit overlooked because of that time pressure. Generative AI helps us do all this critical work—sometimes overlooked—within the timeframe we have.

Caroline Goulard:
Talking about code—do you code quicker with generative AI? Because this is a bit of a myth.

Robin Doumerc:
Absolutely. I don’t think it’s a myth; I think it’s a reality. We do code quicker, especially going from the idea to the prototype. This is now extremely quick.
We can have some ideas that used to take a few days to turn into a working prototype—now it’s a matter of minutes.

Caroline Goulard:
And do you see the same kind of thing on your clients’ projects or within their organizations?

Robin Doumerc:
Sometimes. However, usually in our client projects, the code aspect is just a small piece of the puzzle. So even if the code goes faster, there are a lot of organizational and communication challenges. This is what actually makes a real product. These challenges are not overcome by generative AI. But the software piece definitely goes quicker, and coding faster helps us iterate faster. So we can also improve and have more communication.

Caroline Goulard:
And do you see some organizational challenges not only in how the project is structured or progressing, but also in training people on those new generative AI skills?

Robin Doumerc:
Yes, absolutely. Training is a key part of generative AI because it’s a very powerful tool in the hands of everyone. However, there are also risks and challenges we need to overcome.
We saw more junior engineers suddenly improve the quality of their code. It’s one of the biggest changes I’ve observed: suddenly, junior engineers are producing very high-quality code. However, we also need to train them on how to really take this code and assimilate it—not just copy-paste code that works.
So there’s a management challenge here: how can we train them to actually make progress and learn? That experience used to come from spending hours and hours in the code. Now they have it at their fingertips. How can we train them to assimilate and not just copy something without the deep thinking they would have had before?

Caroline Goulard:
And so how do you do that? What did you implement in your team to train those juniors?

Robin Doumerc:
Well, we actually don’t train the juniors directly. It’s more a management issue. How do the managers become more aware that this sudden increase in quality might hide something else? How do we reallocate the managers’ time to be more hands-on—not just in project delivery, but in real management functions?
They need to spend more time with junior engineers to correct their code, give them new ideas—not just rely on the machine, but also trust the human experience of the managers.

Caroline Goulard:
Is there a big difference between how the juniors are using generative AI and how the senior engineers are using it?

Robin Doumerc:
Yes, absolutely. Junior engineers are more task-oriented. They have a task to solve, and they just use it like: “I have this problem, give me the solution.”
Senior engineers see it more like they have a buddy engineer next to them. They have ideas, they iterate with the machine, they challenge the machine—“I don’t want to solve the problem this way, can we find another solution?”
It’s more of a conversation for the seniors, whereas juniors are more transactional.

Caroline Goulard:
And this also shows up in their expectations about what they want to do or learn in the project?

Robin Doumerc:
Yes. Juniors want instant answers. They have a problem, they want to solve it. Senior engineers are interested in the craft. They already have an idea of what they want. It’s more: “How can I get there in a more elegant way?” or “How can I make something more durable and scalable for a large-scale project?”

Caroline Goulard:
Would you say the junior engineers you hire are more aware of the latest evolutions and technologies than the seniors? Do you allocate more of the GenAI-related challenges to them because they’re more aware? Or does it not work like that?

Robin Doumerc:
No, it doesn’t work like that. I don’t believe juniors are more aware. Okay, they’re younger and more tech-savvy, but actually I think it’s a bit the opposite.
Juniors know what they know, but they might not have the full depth or know where to look to get proper information. Senior engineers have communication channels—they know what sources to trust.
For example, I use Twitter a lot. I’ve had a list created for 10 years with all the engineers I follow. Senior engineers have broader perspectives. Juniors don’t yet have the experience to know who to trust or where to find information.

Caroline Goulard:
Does that mean everyone on your team has time to keep up to date with these evolutions and train themselves? How do you organize that?

Robin Doumerc:
I don’t know if they all have time, but I tell my team every day: the first thing you need to do when you arrive at work is read.
If you look at my agenda, the first thing in the morning is reading time. Keeping up to date is work. Some people think you have to do it outside work hours—actually, no. It’s part of your job.
Internally, we developed some communication channels—technical reading as a service. Every engineer can say, “Hey, I read this piece, it’s very cool.” That’s one of the guidelines: you need to have read it and found it interesting. No questions asked—just, “This is a really cool piece, you should read it.” That way, we share knowledge, and people don’t spend too much time searching for information—they spend time reading it.

Caroline Goulard:
And what do you see at your clients’ organizations? Is it changing the way they work or the way they define roles inside the company?

Robin Doumerc:
I believe it does, in the same way it happens for us. Generative AI creates a lot of reflection on technical roles—how to handle them, how to train them, and all the challenges that come with it.
We can see it at the client’s place. I think on the coding side, clients really look at how to gain productivity. They usually have massive engineering organizations and want to find the best way to use their engineers more efficiently.
We see them thinking about how to create more tactical, more efficient, maybe even smaller teams that can tackle many different projects—because generative AI has transformative power, and they expect a lot of gains from it. So they’re asking: how can we restructure teams to handle more projects, more efficiently?

Caroline Goulard:
And what would be a good practice that you see in the field?

Robin Doumerc:
I think the best practice is to still give people the power and ability to test things. Don’t immediately go into a restrictive mode with just a few teams. Allow some slack to let innovation happen.
At the same time, be very aware of security risks and how to train teams to protect the organization—like avoiding the disclosure of secrets, for example.

Caroline Goulard:
And the most advanced clients you have—what are they doing compared to the ones who are a bit behind?

Robin Doumerc:
The most advanced ones are those that immediately embraced this technology. Their teams were already innovative, and they adapted their way of working, the software they use, how they code.
Some teams immediately saw this as a gift: “How can we take this and use it?” These teams learn fast, share knowledge a lot—communication is key. They share best practices across the organization.
Less mature teams were more risk-averse at the beginning. They wanted to regulate usage before giving access to everyone. So usually, only a selected few had access to generative AI, and knowledge transmission was slower in these organizations.

Caroline Goulard:
Would you say that generative AI creates a very big gap compared to previous developments in AI and data?

Robin Doumerc:
I believe it does, in the sense that AI used to be reserved for the happy few. You had to be an engineer to harness the power of AI. What’s changed with generative AI is that suddenly, everyone can use it. You don’t need to be an engineer, you don’t need to be a strong coder, or even have machine learning studies. Anyone can use the tool, talk to it, and improve. It’s open to everyone, and everyone can really progress because of that.

Caroline Goulard:
And this difference in terms of maturity—you explained it’s important for knowledge sharing. But do you also see a difference in terms of profiles or job descriptions?

Robin Doumerc:
I don’t think some roles will disappear and new roles will emerge. I don’t think it’s going to be a massive change like when AI first came and we started creating roles like data scientists. However, responsibilities might evolve. Classical AI will still exist and still be used. We’ll still need the same profiles.
What might shift is the focus—less on code writing, more on business understanding and solving business problems with AI in a general sense. Engineers will still be needed to build pipelines, frontends, etc.—it’s just that some of their time will be reallocated.

Caroline Goulard:
Let’s take a concrete example: if you’re hiring someone and doing interviews, is there something new you’re looking for that you weren’t before?

Robin Doumerc:
The way we do interviews has changed tremendously. It used to be easy in technical roles: give a simple coding interview, ask them to solve a problem. Now, these kinds of interviews are pointless—generative AI will solve them better.
So we’ve shifted. Now I focus more on engineers, not just coders. I want to see how they tackle business problems, how they bring value to an organization—not just how they solve a purely technical problem.

Caroline Goulard:
Are you looking for different skills today than before?

Robin Doumerc:
No, it’s roughly the same. We want people who can formulate a problem, express it, find solutions, and sometimes come up with different options.
It could be the easy one, the more complex one—then figure out which is optimal: most efficient, quickest to implement. These were always the skills I was looking for.

Caroline Goulard:
What’s the perfect profile of someone who’s good at that? Someone with a lot of experience, someone with the perfect CV or the right school, someone who codes a lot? What are you looking at?

Robin Doumerc:
The main characteristic I’m looking for is curiosity. The field is evolving extremely rapidly—every week or month there’s new technology, new everything. I need someone who’s very interested in these changes, who isn’t afraid of challenging themselves.
Someone who says: “I knew this, but now it’s changing—let’s discover what’s next.” That’s key for technical people: curiosity to keep learning and challenging themselves.

Caroline Goulard:
There are massive amounts of announcements and changes in tech. How do you know what’s really game-changing versus incremental?

Robin Doumerc:
Sometimes, there are big changes, but often they’re just incremental—not groundbreaking. You need to keep a cool head about everything happening.

Caroline Goulard:
So, you’re still on Twitter?

Robin Doumerc:
I’m still on Twitter. That might change, but for the technical side of things, a lot of engineers from Silicon Valley are really on top of it.
There’s a whole new field around multimodality—combining images and text. We’re exploring how to give images to AI, how to interpret them, and how it unlocks new capabilities.
Before, handling images, talking with them, extracting content—was hard. Now there’s a real breakthrough with visual language models. It’s very exciting.

Caroline Goulard:
Do you have plans to test this?

Robin Doumerc:
Yes, absolutely. It could unlock more time for people to work on higher-value tasks.

Caroline Goulard:
And I imagine you sometimes meet other CTOs from other companies. Do they have the same challenges in mind, or do they have specific ones?

Robin Doumerc:
I believe they share some challenges. But one of the big ones I see is when you have a very established organization, you usually have legacy software, with some systems running for 10–15 years.
Sometimes, the people who built this infrastructure are no longer there. So there’s a trend in what I see: it’s all about knowledge sharing and documentation.

Caroline Goulard:
So for you, is the real impact of generative AI about information?

Robin Doumerc:
I believe so. It can be about sharing or extracting it. But at the end of the day, we all progress collectively if we communicate. It’s about how we share knowledge to make everyone aware and help them progress.

Caroline Goulard:
How do you see the future? There’s that quote from the CEO of Nvidia, that IT will become the HR of AI agents…do you think it will happen?

Robin Doumerc:
I think there’s still a bit of hype there, but I see why it’s being said. It’s a bit cheeky—but I get it. IT used to be considered a cost center. Now, with generative AI, it suddenly becomes a profit center.
I think there’ll be a lot of reorganization—IT will work more closely with the business, to be part of the change and help solve business challenges.

Caroline Goulard:
To conclude, do you have a recommendation for people who want to go deeper into this topic—a book or article, maybe?

Robin Doumerc:
Yes, there’s a great newsletter called The Pragmatic Engineer. It’s from a former engineering manager at Uber. I really like it because he goes deep into demystifying aspects of engineering.

Caroline Goulard:
See you very soon for a new episode of The Bridge.