IA agentique

Gartner's 2025 Hype Cycle for generative AI defines IA agentique as a system that perceives, decides and acts autonomously or semi-autonomously to reach a goal, and calls it a fundamental shift away from chatbots passifs. The definition is settled. The consequences are not.

Aiman Ezzat, directeur général of Capgemini, told Investir in July 2026 that IA générative delivered individual productivity gains of limited impact, while agents execute tasks, insert themselves into business processes and contribute to decisions. He puts numbers behind it: more than 30% acceleration of application development on some projects, close to 20% fewer incidents and service interruptions, and generative plus agentic projects rising from 6% to over 11% of first-quarter bookings. He also concedes the hard part, "nous sommes très loin du plug and play".

The displacement runs wider than delivery. Philippe Ensarguet argues that the traditional route to architect, years of writing code, is being absorbed by AI at the moment platformisation raises demand for architects. Bain & Company sizes the cross-system coordination work at roughly $100 billion in the United States, with 90% still uncaptured. Mathieu Grymonprez of Adeo expects an avalanche of requests rather than fewer developers: tickets that used to sit at P10 become P2.

MIT NANDA found 95% of enterprise AI pilots delivered no measurable P&L impact, and named agentic systems as the phase that follows. Nothing in that sequence says why the next wave converts better than the last.