Jean-Philippe Faure does not mince words when they discuss the importance of education in the way Eiffage deploys a generative AI in its European construction and infrastructure activities of 85,000 people. The IOC of the income group of 23 billion euros says:
It's not a game, it's a tool. I tell people that it will probably be the 80 minutes of the most precious training in your whole life.
This 80-minute training refers to the compulsory e-learning course that each Eiffage employee must follow before having access to the company's internal internal internal generative intertestation system, propelled by Google Cloud. It is an approach that is at the heart of the Eiffage AI strategy, focused on the supply of measurable value rather than technology for technology. He explains:
We could have said that everyone has free access. But with free access, you can measure the cost, but not the value. These are two different things.
This emphasis put on structured and value -focused implementation characterized the Eiffage approach in terms of AI since it started serious experimentation two years ago, leading to a strategic partnership with Google Cloud announced at the end of 2025.
According to Faure, Eiffage's journey began with an understanding that without data, AI is fundamentally useless.
If you have no data, AI is very difficult,
This has led to the creation of a data center whose company previously lacked, establishing the foundations for its new AI generative initiatives. He says:
Two years ago, many companies sought data scientists – if you did not have data scientists, you were considered a deceased company. My position was that I did not know how we could feed these people due to the lack of data.
The arrival of AI generative technology has changed Faure's reflection on how the company could approach its strategy:
Fortunately for us, it is at this point that the generator arrived – it was an important movement. In this document, our strategy was AI, but we have changed for a generative AI because technology is so sophisticated, so powerful, that it changes everything.
This change led Eiffage to sign with Google Cloud in December 2025, in an agreement that was going beyond the typical-customer supplier relationship. Faure says that what he liked in Google Cloud is the quality of interpersonal relationships. He says:
Our people and people from Google look like twins now – they want to help you, they are not only there to sell you products.
The partnership includes the development of an internal technological platform using Google Cloud solutions such as BigQuery, APIGEE, VERTEX IA and Gemini models. It aligns with What the seller described at his next Google Cloud annual conference in Las Vegas last weekWhere it aims to position itself as an orchestrator of corporate ia ecosystems, emphasizing interoperability and choice.
Likewise, Eiffage focuses on the diversification of its technological succession. In recent years, the construction giant has advanced a multi-cloud approach to avoid too much: any seller. Faure explains:
It was also important for us not to put all our eggs in the same basket – we have a large basket with Microsoft and if we put all the AI with Microsoft, it would be an even larger basket. So we thought, let's balance that.
User cases led by the company
Rather than trying to implement hundreds of cases of use of AI simultaneously, Eiffage has adopted a more measured approach, selecting only 15 applications with high impact to provide in a year – all coordinates and validated by professional users. Faure explains:
I have colleagues from other companies that have 200, 300,500 use cases – but I did not have the money or the resources to provide so many use cases. So we said: let us bring the business, accept this, let's work together, and they all decided on these 15 use cases.
From December 2025 to May 2025, Eiffage focused on establishing the Cloud and Data Foundation, using these four months to “connect all points between Google Cloud and Eiffage”. This resulted in an AI AI event in early May which brought together teams from the whole company. This included around 40 employees of the sales teams and 10 IT employees.
A key aspect of Eiffage's approach has been to establish clear basic rules on data ownership – where the company is in the foreground. Faure says:
We have created rules, where the company is responsible for their data. I am not responsible for it. If your data is shit, how am I supposed to know? There are many tools within the group, but you are responsible for your data – don't blame me. Balk on.
The mandate of online learning
The most interesting element of the implementation by eiffage of the generative AI is perhaps the requirement that employees take an online learning course of 80 minutes – developed by the company itself – before having access to its secure generative AI system. This approach, approved by the CEO and the CFO, creates an exchange of useful value: an investment in employees to access the tools. Faure says:
Our CEO and CFO both agreed that to access our internal generative AI, which is a secure generative AI, you must take an online learning course. It's free – once you've done this, you have full access for free. If you don't take the course, you don't have access. Period.
The justification goes beyond the simple door. Faure considers training as essential to ensure that employees can extract a maximum value of technology:
For me, it is logical to learn to invite. The difference between basic incentive to high -level incentive – what I call incentive with reflection – the output is fundamentally different.
This choice reflects mature recognition that so that the implementations of corporate AI succeed, technology alone is not enough – the development of user capacity is just as important. The training emphasizes techniques such as reflective incentives, where users ask the AI to suggest ways to improve their prompts. Faure adds:
For example, you can ask the AI, could you ask me questions to make sure that my prompt could be even better? Because I know what I try to ask, but it may not be clear for AI.
Early results
Although still in the first stages, the Eiffage approach already shows signs of success. The company has trained more than 2,500 employees on its generative AI system, all these employees participating in online learning, with use data providing information on how technology is applied in the company. Faure says:
We measure the value of the adoption rate. I know that all the prompts, I have a lot of data – 40%of the prompts are for decision -making, the coding is 5%, the office tool assistant is 5%, helping to write e -mails is 5%, the management of the text is 40%.
A notable success came from an engineer who used technology to considerably reduce the time necessary to analyze requests for proposals (DPS) for potential customers in the public sector (a key area for Eiffage). He adds:
One of the guys said they had to answer a DP but that it takes a long time. Using a generative AI, the DP reading part has gone from two days to twenty minutes.
The value of this approach led by the company is that experts in the field can create prompts that IT specialists, such as Faure, could not create:
I could not write this prompt, because I did not know what they were doing – they know what they are doing. He therefore wrote him invites him by himself. Now we can take the prompt, take it to the “gymnasium”, the train, and now the invitation is even better.
This collaborative approach has enabled the company to create an effective library of prompts that can be shared between different commercial units. It was the hope of Faure, because it focuses mainly on the delivery of use cases which benefit from a transversal advantage:
It was possible to take this prompt and make it available for all people. What you are, whatever the work you do, you can access all the prompts. Now energy people, not just construction, can use it. This is common to the whole business.
Faure is also aware that people were already using a generative AI throughout the business, but the strategy was not aligned and the tools used were not secure. Rather than restricting people, Eiffage wants to create an environment where experimentation is encouraged – although as long as education and appropriate systems are used. He says:
Some people used Chatppt – and I don't want to say no, no, you can't use it, then block the Chatppt. I didn't want to do it before I have an alternative to Chatgpt. Now, I have an alternative, I will move everyone from all other generative AI options to the generative group of group.
Has as a personal assistant
Faure takes care to position AI as an increase in human capacities rather than a replacement, which is particularly important in France, a country that has very strong labor laws. He says:
The generative will not replace my people, but now they all have a personal assistant. In terms of work, I will not be able to increase the number of people, but I will have my more efficient people, bringing more value.
And when the CEO and financial director of Eiffage asked questions about the return on investment, Faure adopted a refreshing and honest approach:
I told them, I won't lie to you, I don't know the value. But I have to be intelligent and work with businessmen in order to get value.
Instead of making promises on specific financial yields, Eiffage measures value through adoption rates and tangible time saving in business processes. The example of DP, which reduced a process from two days to 20 minutes, represents the type of concrete improvement it uses to justify the investment.
With the initial implementation showing promising results, Faure is now focusing on the adoption of the organization:
I want to speed up. We know it works.
