How SFR turned AI training into a strategic channel to surface use cases ready for scale

"One of the most appreciated trainings at SFR for a long time."

Training Direction, SFR

The comfortable compromise SFR refused

The AI training industry in 2024 rested on a comfortable compromise: teach employees to use Copilot, three days, back to the office, box checked. This approach transformed nothing. It only let many write in their reports that their teams were AI-trained.

At SFR, the head of skills development and training, alongside the IT modernization director, refused this arrangement. Microsoft Copilot was the official tool. Impact, the internal Gemini-based tool developed by the IT teams, was being deployed. And yet, like in any 8,000-employee organization, shadow IT existed: ChatGPT consulted on the side, sometimes with sensitive data that should not have entered it.

Their brief was not to train on Copilot. It was: train teams on generative AI by facing field reality (including employee ChatGPT use, to teach them to secure it rather than deny it), surface concrete use cases in training by the teams themselves, business unit by business unit, ready for the IT team to industrialize in Impact, and hold a level of demand commensurate with SFR standards over time.

This unusual brief tipped the Catalia mission since December 2024.

Note: The full English version of this case study is being finalized. For the complete content with detailed methodology, results, and lessons learned, refer to the French version.