BOUYGUES TELECOM launches its next-generation AI sales assistant

Bouygues Telecom is revolutionizing the customer experience by launching the first generative AI-powered personalized mobile sales assistant in Europe, built on Google Cloud. Supported by Artefact for over a year, the telecom operator is enhancing its online sales journeys by offering 24/7 assistance to prospects and customers.

Creating an always-on AI agent for mobile offer sales

The increasing number of online mobile offers—combinations of handsets, plans, and financing options—can make it challenging for users to understand, sort, and choose. Until now, online support services had limited availability and capabilities, often leading to frustration and drop-offs, especially for more complex offers. To address this, Bouygues Telecom’s digital team, Lead Open Innovation, initiated a transformation of the support experience using Generative AI.

Three main goals guided the development of this next-generation chatbot:

• Provide 24/7 continuous assistance;
• Deliver a more user-centric e-commerce experience;
• Develop a leadership position in innovation.

A technological breakthrough with Google Conversational Agent

In September 2024, Bouygues Telecom reached a major milestone by becoming the first European telecom operator to launch a generative AI-based sales assistant based on Google’s Agent Builder technologies. This achievement stems from a three-way collaboration between the operator, Google Cloud, and Artefact. with Google’s agent builder technologies

The development unfolded in two complementary phases. In early 2024, the team focused on building the assistant using Google’s Agent Builder, a low-code platform allowing business teams to prototype and test quickly while providing a user-facing interface. With just two technical and two business profiles, the team leveraged Gemini AI to create the conversational agent.

“Agent Builder is a low-code solution that allows business teams to be truly involved in building, testing, and deploying these agent-based solutions.” François Gette, Lead Open Innovation, Bouygues Telecom.

In early 2025, Artefact supported the development of advanced scoring tools to evaluate the quality of interactions between users and the AI assistant. The goal: to ensure continuous user experience improvements using the open-source Langfuse tool.

This first phase targeted prospects—visitors without a Bouygues Telecom account to simplify deployment while addressing acquisition needs.

A chatbot that acts like a real salesperson

The innovation lies in moving away from traditional rule-based chatbots. This next-gen assistant handles natural conversations, digressions, multi-criteria comparisons, and provides tailored recommendations based on a prospect’s budget and preferences.

The assistant goes beyond conversation: The user starts by choosing a phone model, then selects the color and memory capacity. The assistant asks for further details to build a full offer. It synchronizes real-time chat with website navigation to display product pages. Once the phone is selected, the assistant guides the user through choosing a plan and financing method—again syncing the chat preferences with the website selections.

“We were able to convert this behavior into a prompt for the Vertex Agent Builder. Early testing showed real potential.”
François Gette, Lead Open Innovation, Bouygues Telecom.

The assistant supports users through every step: identifying needs and criteria, recommending phones within budget, selecting technical options, choosing a plan and financing, and finally guiding them to the cart, just like in-store assistance.

The Google Cloud ecosystem driving performance

The solution relies on a robust technical architecture fully hosted on Google Cloud Platform, whose core services include:

• Conversational Agent: AI assistant engine powered by Gemini;
• Dialogflow Messenger: integrated customer interface;
• Piano Analytics: performance tracking;
• Langfuse: GenAI and agentic application monitoring tool.

Data quality is crucial to avoid AI hallucinations, especially in a commercial context where accurate prices and product information are essential. The team prioritized qualified APIs that provide real-time product and availability data. This approach not only improves reliability but also drives a virtuous cycle of internal data infrastructure improvement.

A success built on collaborative and agile methods

This project illustrates the power of innovative collaboration between all three partners. Bouygues Telecom’s team effectively used Gemini to generate the assistant’s own instruction set, allowing the AI to self-optimize and enhance performance.

The agile development process also proved successful, thanks to continuous testing with Google’s engineering teams using beta versions of the tools. This collaboration enabled exploration of the technology’s limits while benefiting from first-hand design expertise.

This innovation in user experience is reflected in the real-time synchronization between chat and web navigation, complemented by a unified, user-focused interface for both desktop and mobile.

What’s next: Transforming the mobile experience

Bouygues Telecom plans to expand this initial success. The company intends to deploy more next-gen conversational agents, strengthen a dedicated team for continuous improvement, enrich the AI assistant with new prompts, and develop an even richer UX for a premium customer experience.

“The mobile experience is a bit different, as the chat window is your only interface. We envision a future where mobile interaction is entirely driven by chat assistants. It could become a new channel for product sales and distribution.” François Gette, Lead Open Innovation, Bouygues Telecom.