Laurent Aufils argues Orange's next competitive advantage is knowing when not to use AI, because thinking builds judgment.
Orange's agentic AI saves 40% on troubleshooting tickets, but savings don't translate one-to-one as tickets get more complex.
Orange's U Time gives every employee three hours for training, and AI ambassadors get recognition in reviews.
Knowing when not to use AI is the next advantage
Orange is running an AI transformation inside a very large telecom operator, and Aufils owns both the AI roadmap and the people side. His approach is to analyse work processes first and decide where agentic AI adds value and where human judgment differentiates. Saved time is shared between company, employees and clients, and the harder discipline is knowing when not to use AI.
Takeaways
Transformation lives in jobs and workflows. Orange's bias as a tech company is to start with the tool, but Aufils says success depends on how AI sits in jobs, workflows and processes. Adoption paces differ, with some employees excited and others anxious about their jobs.
Ticket AI saves 40% of the time. Agentic AI cuts the time to treat network troubleshooting tickets by about 40%, but the saving does not convert one-to-one into 40% more tickets. Customers ask for more information, and the tickets left are more complex.
Saved time is split three ways. Orange's principle is that saved time produces dividends balanced between the company, employees and clients. Clients can get extra services, and the company learns from tickets to fix root causes. Employees get learning built into the working day.
Theoretical productivity gains rarely survive implementation. Implementation creates new tasks such as supervising and adjusting what an agent delivers, so actual gains lag promised ones. Aufils also pushes back on the bias that AI equals productivity, since productivity means doing more with less, not growing.
Scaling AI will raise token costs. Putting AI into everyone's core daily activities will mean overall token consumption far higher than today, especially if unit costs rise. Aufils places this in a new maturity stage after the first phase of drafting documents faster.
Ambassadors beat mandatory training. Orange's volunteer AI ambassadors run for about a year and their contribution is recognised in yearly performance reviews. The programme surfaced AI enthusiasts from outside IT, and Aufils argues peer example is more powerful than mandatory training.
Reasoning through problems builds judgment. Aufils says the key question for the future of work is when not to use AI, and that reasoning through a problem builds judgment. AI cannot take responsibility for choices, so Orange keeps hiring graduates to build distance from AI output.
Where humans stay in customer care
Some clients want a fast answer and are comfortable with a bot.
Others need a person who knows their context, which matters more than speed.
Agentic AI at Orange often equips the employee taking the call rather than facing the customer.
Tap a name to go there · tap a line to see why they are linked
On this episode 2
Show 1
Relationship building happens in the customer conversation, so it is lost when that moves to a tool.
The warning to HR
AI is a chance for HR to redesign its strategic contribution.
Without a seat at the strategic table, HR risks being reduced to execution work AI can take over, which Aufils calls an existential question for the function.