From capability to agency – when AI starts acting on your behalf

Two years ago, we explored how not to get fooled by AI. The conclusion: Capability alone isn’t enough. AI needs knowledge, reasoning, oversight and trust. Today, AI is moving beyond answering questions to planning, deciding and acting. We’re entering the age of agentic systems. The challenge is no longer just intelligence, but authority: from copilots to autonomous agents, from prompts to decisions, from capability to accountability, from automation to delegated authority. ASML’s Alexey Nesterenko brings this challenge into engineering reality, transforming complex software-change documentation from a growing “big book” into AI-powered answers. The experience shows that AI itself is often the easy part. Good data, context, understanding ambiguous engineering language and human oversight are what make it work. Start small, learn from real users and let the solution evolve, while keeping humans in control. These lessons become even more important as AI moves from answering to acting. Robert Engels explores how organizations can embrace this shift without losing control. Because the defining question is no longer: “Can AI do this?” but “Who gave it permission?” 2024: Don’t get fooled by AI. 2026: Don’t lose control of AI.