Agentic AI for large organisations: beyond token-maxxing

Published by BNP Paribas CIB, based on Charles’s keynote at the Mistral AI Now Summit in Paris. Includes the full video of the presentation.

Key ideas

• Beyond token-maxxing. Success with AI is measured by business outcomes — revenue, productivity, customer impact — not by the volume of tokens consumed.
• From Generative to Agentic AI. Generative AI amplifies individual tasks; agentic AI “rewires” end-to-end workflows through autonomous agents that reason, act and collaborate under human supervision.
• Trust is the differentiator. With the same foundation models available to everyone, advantage comes from grounding agents in governed, proprietary data products and expert-designed workflows.
• A six-pillar framework for moving from pilots to enterprise scale: business-led strategy, technology orchestration, data as a product, automated governance, people and culture, and FinOps with a “value-per-token” lens.

“If what got us here won’t get us there, we must rewire the ecosystem from the ground up.”

“For our corporate and institutional clients, the transition to agentic AI is a structural evolution of their operating models. It requires strategic investment and a long-term view.”