Orange employees have integrated generative AI tools into their professional routines shortly after their emergence. Through acculturation and hands-on training for all, coupled with proactive communication, the adoption of generative AI is set to become a lever for efficiency and innovation among the Group's employees.
A First Step into the World of Generative AI with Dinootoo
Dinootoo is an internal generative AI tool at Orange that brings together several Large Language Models capable of meeting the specific needs of employees, particularly in decision-making, while ensuring data security.
In just one year, more than 50,000 employees have adopted this tool, representing almost half of the 110,000 eligible employees. An internal study showed that the most active users save an average of 2 hours per week. Employees have embraced certain use cases, such as summarizing meeting transcripts and documents, rephrasing writing, image analysis, and document drafting assistance. This also includes tasks in legal, procurement, programming, and agile coaching, among others. Between 10,000 and 12,000 employees use Dinootoo each week, and these numbers continue to grow, notes Cédric Goutard, Team Leader and Product Owner of Dinootoo.
However, some employees still face barriers to its use. Philippe Trimborn, Director of the Institute of Professions and School Relations at Orange, points out: "Some people do not understand what it will bring to their work and have a real reluctance to engage in interaction with a machine." It is therefore essential to support and engage employees to help them understand Dinootoo and the benefits they can derive from machine learning.
The training sessions organized by Orange have played a crucial role in this adoption, with more than 35,000 employees trained in less than a year, helping them familiarize themselves with the capabilities of generative AI and understand how to use it in their daily work, job by job.
Some updates have also facilitated the adoption of the tool. For example, the action library helps users discover available features through predefined prompts, while allowing them to select options based on their work habits.
Constantly evolving, with its computing power and massive data management, Dinootoo will gradually be enriched with new models and specific prompts by job, to better meet the diverse expectations of employees.
Lessons from Generative AI Experiments such as Copilot for Microsoft 365 and Github Copilot
In parallel with the development of Dinootoo, Orange has also tested several existing generative AI tools.
Copilot for Microsoft 365 is one of them. The experiment, conducted with 3,000 employees from various countries, entities, and functions, showed that adopting this solution is not as simple as its integration into office software might suggest. The average time savings of about 1 to 1.15 hours per week reported by testers is "real but remains moderate" and varies depending on the jobs and tasks performed, says Marie-Hélène Briens Ware, who leads the development of Employee Experience topics for Orange Business clients. For example, it allows for the automatic creation of summaries from meeting transcripts and recordings on Teams. It also helps draft and summarize emails, which is particularly beneficial for jobs where urgent and important exchanges are frequent, such as in the B2B customer service. Directors in this service use Copilot to become aware of issues that arise, respond more quickly to urgent emails, and make informed decisions. Additionally, Copilot facilitates sales and pre-sales processes by assisting in drafting business proposals and managing communications. Emphasizing the need for support to help employees adopt the tool is also crucial.
Among the barriers to using Copilot for Microsoft 365, the cost of the tool, which is around €30 per user per month, raises questions about return on investment, as well as the complexity of implementing effective data governance, requiring special attention to ensure the security and quality of the information used. But for Marie-Hélène Briens Ware, integrating generative AI directly into daily work tools can offer personalized and proactive solutions in the future, a potential that seems significant with Copilot for Microsoft 365. It is therefore important to continue supporting Copilot M365 users and addressing concerns related to data governance and carbon footprint, she says.
Simultaneously, the experiment with Github Copilot, an AI capable of increasing developers' productivity, was conducted. The promise of this tool is to write code faster, with less effort, and for a better quality result. At Orange, Sofrecom participated in a test phase in 2024, involving a sample of volunteer employees from various development professions. After nine weeks of experimentation, developers noticed an increase in successful compilations and a reduction in code anomalies. Overall, productivity gains were estimated at between 10 and 15%, and a notable improvement in code quality was observed. Mohamed El Mrabti, Department Director at Sofrecom Morocco, specifies that "experienced developers particularly appreciated the tool, while less experienced ones needed more support to fully benefit from it." This echoes another observation made by Philippe Trimborn: "the adoption of generative AI is influenced by the level of education, with a positive correlation between qualification and the use of these tools." These feedbacks highlight the need to adapt support to different levels of experience and competence.
For 2025, the challenge will be to "train and support everyone in relevant and responsible uses that will generate efficiency gains, as close as possible to their job gestures, relying in particular on ambassador communities," explains Roxane Marsan, Director of Orange Campus Tech. "And this includes developing our critical thinking on biases and hallucinations, ethical and environmental issues."
In summary, the integration of generative AI tools such as Dinootoo, Copilot for Microsoft 365, and Github Copilot by Orange demonstrates the group's desire to encourage employees to familiarize themselves with the uses of generative AI to understand its full potential. Although barriers remain, particularly in terms of understanding and adoption, the observed productivity gains indicate undeniable potential to transform employees' daily work.
This is why this transformation cannot happen without appropriate support, taking into account the reluctance to use it and the various skill levels of users. Thus, Orange will face new challenges, particularly in terms of data governance and carbon footprint, while adopting best practices related to these technologies.
