IDeepTalk with Marcelo Goes CIO Data & AI at L’Oréal

How can AI move from experimentation to deployment at scale?

The figures Marcelo shares convey the scale of the effort: 73,000 employees trained in AI, 220,000 hours of training in 2025, and 25,000 unique users of L’Oréal GPT every day. The group also analyses 30 million consumer conversations daily, while more than 80% of its applications run in the cloud.

Moving from experiments to a shared approach

For Marcelo Goes, technology alone is no longer enough. The real challenges lie in data quality, security, processes that work across teams, and supporting people through change. As use cases multiply, scaling them requires a common framework.

L’Oréal has brought agents, skills, MCP and APIs together on a single platform, where these resources can be catalogued and managed. The aim is to move beyond isolated proofs of concept and enable teams to develop AI use cases at scale within a controlled environment.

This approach also rests on a principle Marcelo sums up as: “No master data, no data. No data, no AI.” Maintaining data quality requires continuous attention. The choice of model, meanwhile, depends on the use case: Gemini, Claude, OpenAI and Mistral can each serve different needs. The episode also explores L’Oréal’s recently announced partnership with OpenAI.

A transformation centred on people

Training plays a central role in this transformation. Marcelo describes AI as a way to enhance employees’ capabilities, rather than replace them. He points to customer service as one example, citing an 80% productivity gain on the tasks concerned alongside a 10% increase in satisfaction.

The balance between performance, support for employees and risk management echoes the discussions taking place within Positive AI. In this episode, he explains how exchanges between companies help address challenges that continue to evolve, from bias and environmental impact to sovereignty and the potential misuse of AI.

A conversation offering a practical look at what it takes to deploy AI across a large organisation, and the value of sharing experience among practitioners.