Orange is pushing deeper into enterprise AI, but with visible caution.
The operator sees major financial value in automation and smarter workflows. Yet it also wants humans to remain in control. That balance now shapes its AI strategy across Orange Business Services.
Miguel Alvarez, chief data and AI officer at OBS, rejects the idea of workers serving machines. He referenced Cory Doctorow’s “reverse centaur” warning, where AI guides humans instead.
“I think it’s a very good metaphor,” said Alvarez.
For telecom teams, that concern feels practical. AI can speed coding, product design, and contract checks. However, it can also create errors at scale. That risk grows when systems act with less human supervision.
Orange says its strongest AI gains do not come from cutting staff. Instead, teams shorten development cycles and protect revenue. They also avoid costs and launch services faster.
OBS reached a €50 million value creation target in 2025. This year, the goal rose to €70 million. At group level, Orange aims for hundreds of millions of euros.
The company now explores several hundred AI use cases. Around 30 already run in production at OBS. Another 20 are moving into broader deployment.
Adoption has also exceeded early forecasts. Orange first expected 10,000 to 20,000 active users over several years. The company now reports more than 100,000 active users.
One example involved a new product development project. Alvarez said AI helped deliver a working version within months. Without AI, the process could have taken four or five times longer.
Software development has changed quickly too. Some teams still use AI for code suggestions. Others now use AI agents across larger development tasks. These agents can interpret requirements and automate several steps.
Still, Orange does not want uncontrolled automation. Alvarez warned that AI systems often behave unpredictably. That makes guardrails essential, especially in business-critical telecom environments.
The company also limits dependence on external platforms. It uses commercial frontier models when suitable. It also runs open-weight models on its own infrastructure.
That approach matters for sensitive contracts and regulated enterprise work. Some data cannot move into public cloud tools. Orange now says on-premises models can handle more demanding tasks.
Alvarez described the newer balance clearly.
“We cannot use a cloud-based model to check – a frontier model to check – because for some contracts that’s not appropriate,” said Alvarez.
The broader lesson for telecom operators is direct. AI can create value when business processes change with it. But expensive models alone will not deliver returns.
As Alvarez noted, companies face danger when technology moves faster than transformation. For Orange, the objective remains clear. Use AI as a tool, not as the driver.
