Report at the Matinale de la Data & IA, Paris – Orange bought the Data and artificial intelligence specialist Business & Decision in 2018. An incongruity for a connectivity provider? Not in the opinion of Michael Deheneffe, the consulting firm’s director of innovation.
“Orange’s primary business is Data. Initially, the operator carried voice. It’s been data via fiber, copper, 5G and tomorrow 6G.” Data is everywhere, and in all organizations. A technology is also becoming essential: generative AI.
19% of companies engaged in AI Gen initiatives
Since the launch of ChatGPT, the announcements have followed one another at a steady pace, underlines Mick Levy, director of business innovation. Three illustrations just a few days apart: the next fundraising of 450 million euros from Mistral AI, the release of Gemini by Google (in 3 versions including one on mobile) and a meeting of the European trilogue on AI Act – which drags on by the way.
But generative AI is also a significant interest within organizations (like obstacles). And the Business & Decision executive cites a recent Gartner study: for 70% of decision-makers, this technology is considered to bring multiple opportunities. However, 19% have actually launched AI Gen initiatives, which are often limited in terms of scope and impact.
In fact, only 4% of companies say they are sufficiently mature to begin large-scale deployments. Mick Levy sees a paradox there. “Despite the perceived opportunities, companies are struggling to get there.” It is nevertheless an opportunity, he continues, to have a competitive advantage.
“The first entrants will gain a competitive advantage,” he judges. And the easiest way to start is certainly by tackling uses in the field of document intelligence, one of the three main families of uses along with content generation and interaction (including conversational assistants). .
ChatGPT massive acculturation tool
These types of use are not waterproof. They can be combined, for example by allowing a document database to be queried in natural language via a chatbot. “A call center will be able to automatically generate responses to customer requests,” he further illustrates.
The democratization of AI favored by ChatGPT, widely adopted at the end of 2022 by the general public, facilitates initiatives in collaboration with the professions. “So we discussed ChatGPT on Sunday, as a family, over chicken and fries,” laughs Mick Levy. Data experts gathered during a round table can testify to the benefits of this tool in terms of acculturation.
“My grandmother now understands what my work involves and what AI can enable,” jokes Suez’s Head of Data, Claire Mathieu. Other profiles have understood the interest of AI Gen: the members of the Executive Committees and management committees. “They are today demanding to understand the business opportunities, the value potential, as well as the risks,” she declares. The AI buzz of recent months therefore makes it easier, even if other actions have been taken in terms of acculturation.
“Our approach consists of relying on innovation projects, starting small (…) They have generated a lot of value or given satisfaction to operators. The use cases were truly created for operational staff (…) Today, we no longer need to go looking for use cases,” testifies Claire Mathieu.
Creation of an AI committee for Les Echos-Le Parisien
The trend is the same within the Les Echos-Le Parisien group, whose CDO Violette Chomier sits on the Executive Committee. At the start of 2023, the company set up an AI committee integrating all professions and hierarchical levels.
“We listed all the use cases that seemed relevant to us. This also contributed to the identification of other use cases using classic machine learning, in addition to those linked to AI Gen.”
Among these applications is the use of these technologies for software development. “Accelerating developments is a strategic issue for us. Our resources are limited, as is our ability to attract talent,” explains the Chief Data Officer, who also cites the use of AI to improve the accessibility of content (multi-format, multilingual, etc.) and the quality of data. .
Within the mutual insurer Groupama, generative AI is presented as an accelerator of AI projects, as a whole. Classically, Gen AI is used for document research and internal bots, but also for interactions with customers. “A customer use case goes into production in 48 hours,” reveals François-Xavier Enderle, director of group Transformation.
The Groupama method for moving forward quickly
Groupama, however, had to convince compliance and risks. The insurer chose to start and deal with the legal obstacles gradually. A multidisciplinary group called IA Factory was formed, involving compliance from the start of the process. “It’s one of the keys to our speed,” he confides.
“We delivered a first use case in October at group level and two will arrive by the end of the year on the client side,” says François-Xavier Enderle. “We brought all compliance into the AI team, but without putting the brakes on ourselves at the start.” Of course, obstacles remain and adaptations are necessary, for example in terms of contracts with AI suppliers.
Claire Mathieu also points out that generative AI does not constitute the entire project backlog. At Suez, these relate in particular to the sorting of waste in factories and the treatment of water, including the reduction of leaks.
AI at the service of Suez’s CSR strategy
“Households are sorting more and more of their waste and our factories must therefore operate more and more quickly. AI thus intervenes to help us optimize our machines in real time,” illustrates the CDO, which also recalls that “AI serves Suez’s strategy on environmental issues.” Thus, the profession is concerned about the impacts of the AI models deployed.
“The objective remains to protect resources and reduce consumption,” she says. The resource requirements linked to generative AI therefore constitute a problem at Suez, which nevertheless intends to take advantage of it, particularly for the processing of calls for tenders. Let us point out that according to our information, Claire Mathieu also has responsibility for leading the Suez IA Gen program.
Added to these projects are those using machine learning and deep learning, which are not disappearing. Mick Levy encourages taking advantage of the generative AI craze to accelerate artificial intelligence initiatives. “According to Gartner, traditional use cases will increase 1000-fold by 2028.”