Build / Measure / Learn: Our organization to accelerate GenAI deployment

In a context where generative AI initiatives were rapidly accelerating, Sodexo launched a strategic reflection on how to integrate this technology into its activities to simplify the daily work of its operational teams, who are often highly solicited on the ground. The goal: to prototype concrete, high-value use cases, establish the foundations of a clear governance model, and define a trajectory for large-scale deployment.

To support this momentum, Sodexo created a dedicated team in charge of transversal generative AI initiatives, with the mission to structure governance, support business teams and employees in designing relevant solutions, and build on early successes to foster organization-wide adoption.

Within this framework, Sodexo selected Thiga to support the implementation of conversational agents while helping formalize a reproducible and sustainable approach for the entire IT department.
In this context, Sodexo chose Thiga to support the implementation of conversational agents.

• Structure governance around conversational agents: define, prioritize, and secure these emerging projects so they can later be replicated across specific regions or business units.
  
• Strengthen the product approach: integrate Discovery and Delivery best practices so these projects become part of the Product strategy alongside other group initiatives.
  
• De-risk the first prototypes: quickly test the feasibility of diverse use cases while demonstrating added value compared with the use of an LLM such as ChatGPT.
  
• Support the development of AI maturity: integrate into the Group’s AI upskilling strategy to accelerate understanding of uses, potential, and limitations of GenAI, both for Tech and business teams.

This mission enabled Sodexo to build on its generative AI initiatives and strengthen adoption among business teams.

The main outcomes observed:

• A successful experimentation dynamic: two prototypes delivered on time, including the conversational avatar, which generated strong engagement and is set for large-scale deployment across more than 700 sites.
  
• Accelerated learning of Copilot Studio technology: teams strengthened their skills in AI agent design and adopted rapid-iteration best practices.
  
• A GenAI governance model taking shape: definition of a clear framework for scoping, prioritizing, and steering future initiatives.
  
• A reinforced product culture: adoption of a Lean Startup approach centered on user value and impact measurement.
  

.