Fagiuoli explains that the company has used AI and LLMs (large language models) extensively internally for a little more than a year. In both theory and practice, he shares, it’s difficult, natural, and helpful at the same time.
— We are putting a lot of effort into all our R&D activities, where we’ve also integrated AI — particularly in analysing market trends and collecting data on active ingredients. This helps us speed up some phases of evaluation and selection of available technologies on the market. We have a quite large pipeline of areas in which we are testing the AI.
— For the R&D team, the LLM tools can screen all kinds of information and accelerate and facilitate the identification of what is relevant and what is not. We’ve started to integrate some of them into our processes but with caution and a critical view of everything suggested. The key lesson we’ve learned is to never rely solely on AI. It should be seen as an assistant, not a human replacement. Mistakes can happen, so human validation is still crucial. We’ve also learned that AI-generated responses need to be adapted to the customer’s tone and context, which is why full automation is not an option for us.
It might be obvious but can you explain why it’s so important to be critical?
— Because AI doesn’t think. It uses what’s there, and it reshapes the words compared to what is already written. You need to be careful because what sometimes dominates is mainstream thinking. With this mainstream thinking, there is much more content on it, so it orients people in one direction and encourages them to think in the same way. Statistically, you have more content that is on a way of thinking A, and there’s limited content on a way of thinking B. That’s why you need to be critical, and that’s where you sometimes have the limit with these languages.
Which LLMs have you worked with?
— Several different ones, such as OpenAI, Mistral, and Bart, and customised it to make sure that it fits our own grids of analysis.
— Developing large language models is extremely costly — even major corporations struggle to develop their own models — and, as a result, most companies rely on one of the few existing ones. The common approach is to utilise one of these LLMs by integrating it and then customising it to meet specific business needs. That’s exactly what we’ve done. There’s no need to reinvent the wheel, the smartest strategy is to adapt the wheel to suit our objectives.
— Also when L’Oréal launched its own engine, it’s based on one of these technologies. What they did and what we do is helping the team who answer the questions (from customers, Ed’s note) to come with relevant responses. We always have someone who will answer a person but the LLMs help prepare the answers. It’s like if you have a super assistant who provides you drafts, that you can check, review, update, and respond to, which helps to get faster responses to the customers.
And how can your work with AI benefit your clients, both salons and consumers?
— AI mainly helps us by providing standardised and consistent response suggestions for our customer service team. This allows us to work faster, while still having the flexibility to adjust and personalise replies depending on the specific context. It’s making sure that we cover all aspects of R&D, including gaining time, and accelerated response time, plus having relevant answers for customers. When you have a (customer service) team of 5-10 people, how do you train them to make sure that they consistently answer the same way when questions come? Person 1 might be new in the company and is not trained enough, while another one is very experienced, so depending on who’s going to receive the question, the answer might be different. These AI models will accelerate the response time when we have the ’super assistant’ that provides what would be a good answer and makes it easier to reply. It provides this consistency that sometimes companies lack because it’s difficult to train people and make sure that they are all at the same level.
— We are also currently testing AI for prioritising customer service tickets based on urgency or complexity, ensuring that critical issues are resolved quickly. Additionally, we are also in the testing phase of using AI for multilingual interactions with real-time translation, which would allow us to serve customers from different regions more effectively without needing human agents fluent in multiple languages. We are exploring other AI applications as well but it’s too early to go into further details.
The beauty industry often lacks cooperation between brands, can you also work together with other players in order to become better at AI?
— Today it’s more internal. It might be interesting because it’s very tech- and time-consuming and costly so we could benefit from that but as of today, I don’t see us starting to do it with other brands.
— On our side, we’re quite tech-friendly — we know how to do it in an effective and cautious way. Right now, it’s very ’in fashion’, so everybody speaks about it but we use it very cautiously internally. We don’t start claiming to be the ’AI lab’ — that would be nonsense.
Dermaceutic just launched Reveal 4X, the first eye serum in the brand’s range.
— It’s a product that is quite unique compared to what’s on the market, says Fagiuoli. You usually have one, two, or maybe three but not many products that target all four main indications for the eyes at once. I think we found the right balance and the right combination of active ingredients to make sure that each ingredient works effectively on each of these indications without having a negative effect on others. Take caffeine, if you use it too much, it works against puffiness but it can also make the skin too thin afterwards. It has a counter effect on other aspects, so it’s about finding the right balance.
Has AI helped in the product development?
— Yes, in screening and selecting what is out there in terms of technology and making sure that whatever we have selected in the cream is superior to what is already existing on the market.
If you take a future perspective, what do you forecast in AI for you and the industry?
— The two things where I think we will have the strongest processes in automation in place in order to exploit the full extent of the AI technology are within R&D and customer care. It will take some time because people have to change their work habits. What is key for us is to always make sure that between the customer and us, there is always a human interaction. That said, I don’t think we are ready today — or ever — to go through a totally automated process. Today, in my perspective, AI is a good help for assistance. You need to keep it as an assistant and not replace anything with it. People have this belief that AI is clever. It’s not clever. It’s just statistics and tools. Instead, I see it as helpful and it will facilitate the work of the people in the company.
You’ve also stated the importance of having a humble approach to science. Can you elaborate on that?
— When I see the time we spend on the development of a product like the eye serum, we’re talking about years. We always work from scratch, and we try to face the challenges. When we solve an issue and succeed in something, we are very happy — but there are 10 or 20 questions coming out of it. This is a humble approach because you never feel like the final ’Eureka — I found it, that’s the final thing, now it’s finished, and my life is over.’ It’s never like that. One answer provides ten questions. That’s where you need to be humble because the knowledge we have today is nothing compared to what we will know in 10 years.
— The time we spend on things is R&D time. It’s not under the pressure of marketing or sales, saying that we need to launch a product. If that would have been the case, we wouldn’t have come out with this kind of product. Instead, we would have taken a standard base formula, integrated 2 or 3 ingredients that are out there, very fashion and hyped, and launched it in a few months. That’s not our approach. And this is also where we say it’s humble because we think the product we have today, as of today, is the best thing we can get in terms of formulation. In 10 years time, we’ll have a different explanation and may update the formula to be at the edge of what is new.
— When you look at knowledge, the exponential process of discoveries in every field doubles every eight years. How can you even cope with that? This is also why we are humble.
”The knowledge we have today is nothing compared to what we will know in 10 years”
Can you also say that you have a slightly humble approach to AI?
— Yes, exactly. AI is often overhyped, but in reality, we remain cautious in how we use it. It’s essential to treat AI as a complementary tool rather than a miracle solution. We understand that AI isn’t perfect and must constantly evolve. Our approach is measured, recognising its limits and integrating it wisely into our processes.
In what other ways do you think that AI will transform the beauty industry?
— AI has tremendous potential to personalise beauty routines by analysing real-time consumer needs. In the future, I believe we’ll see beauty products become increasingly customised, driven by algorithms capable of recommending treatments tailored to each skin type and specific concerns surpassing today’s basic recommendations. That said, we will always orient the customer to a consultation with a healthcare professional, with whom we work closely. True interaction and expertise cannot be replaced by technology. AI serves as a shortcut, a tool to streamline and enhance the process, but the final step — an informed, human touch — is essential. While some brands may see AI as a way to bypass healthcare professionals, that’s not our philosophy. We see AI as a bridge, not a substitute.
— Today, AI is often used in the beauty industry more as a way to signal that companies are at the forefront of innovation, rather than reflecting real, tangible integration. Many players are quick to communicate about AI, but the actual impact remains limited. It’s a slow process. Large corporations are structured with deeply ingrained work habits, making it difficult to implement these new tools effectively. While communication around AI can happen quickly — it only requires a few key buzzwords and some articles — the real integration will take years. This is why I approach the topic with humility. AI is hyped right now, but we must acknowledge that it’s a gradual evolution, not an overnight revolution.
2025 is approaching, what are you looking at?
— We have a nice pipeline of product launches in the next three years. I think Dermaceutic today compared to Dermaceutic in two, or three years will be totally different. In 2025, we are in a little bit of transition. We will extend our product portfolio progressively with one or two launches every year between home care and professional, and some upgrades of existing products, integrating things we have seen that are relevant, consistent, and sustainable in the long term in the existing formula.
”We must acknowledge that it’s a gradual evolution, not an overnight revolution”
Except for AI, one major macro trend now is the rise of wellness. How can you cater to that, if you can?
— One of the three pillars of science is about patient indication and concern. And we see that people are both very savvy and early about all this, and more and more younger people are conscious about their health and living a healthy life. This means that we need to have answers and solutions for people — even younger people — which means softer, progressive approaches and protocols. It’s also about explaining that you could have small treatments all your life to maintain, and to avoid doing these big jumps, such as getting a strong peel just because you’ve been through an overwhelming, too-extreme life.
— With people being very conscious about this and wanting to have something that is a bit softer yet effective in targeting the indications they want, we’ll probably see new products that will address this kind of need. It’s about having a lighter and softer approach that gets good results over time, and good aftercare.
And the products are just getting more and more advanced.
— Today, people deserve more explanations. You have a lot of brands out there, and claims are sometimes quite similar from one brand to another. Active ingredients are sometimes quite similar, and what comes after is a bit of education and understanding of what’s behind the product. We’re getting into a high sophistication compared to a few years ago. When people have difficulties in understanding the diversity in terms of an offering, certain explanations and expertise need to be communicated to them. To stay relevant, we need to deepen down a little bit and explain more about our reassurance to do our best in terms of what you can get on the market, says Fagiuoli. He adds:
— It’s also important to highlight the importance of transparency in this new era of AI and science applied to beauty. As a brand, we strive to maintain an honest and rigorous approach, explaining the scientific truths behind each innovation. Consumers are increasingly demanding proof and results — and it’s our responsibility to provide that clarity!
