How Bouygues Telecom Built AI Agent Iris to Give 6,000 Reps Accurate Answers in 30 Seconds

Bouygues Telecom is one of France's three major telecom operators, competing in a market defined by rising customer expectations and shrinking margins. In a commoditized industry, service quality is the primary differentiator — and Bouygues has built its strategy around putting AI at the hands of every customer-facing employee.

The core challenge was response quality at scale. Across 6,000 contact center reps and 500 retail stores, the speed and accuracy of answers to customer questions varied. Reps had to manually search through systems to pull relevant account history and service data, which meant customers waited minutes for information that should take seconds — and the quality of the answer depended on the rep's experience.

Bouygues built Iris, an AI service agent powered by Salesforce Agentforce and grounded in more than 500 help articles and live service data unified through Data Cloud. When a rep receives a customer query, Iris surfaces a precise, tailored answer in seconds — drawing on account history, case context, and real-time service data to construct a response specific to that customer, not a generic FAQ. The same agent generates pre-call context summaries so reps arrive at every interaction already briefed, and drafts post-call follow-up texts that go out at scale: 10,000 per day across the contact center.

The 6,000 reps and 500 retail associates now work with Iris as a permanent layer in their workflow. Iris achieves 95% answer accuracy, and adoption is voluntary but near-universal: 90% of service reps rate Iris 4 or 5 stars. The model is consistent across the Parisian customer service floor and in-store in every city Bouygues operates.

Bouygues is now extending the same Agentforce model to its 10,000 field technicians via Agentforce Field Service and Operations — providing context before each job, guidance during it, and structured follow-up after it. The target: 10,000 more completed service appointments per day without growing the technician workforce, by reducing the time each tech spends on non-technical coordination before arriving at a site.