I just finished reading the Gartner report (Emerging Tech: Agentic AI Maturity Roadmap) If I understand the message correctly, agentic AI is emerging as the next major promise. I have no doubt that organizations will embark on this path. But I am convinced that significant setbacks will be encountered.
Agentic AI: A Non-Deterministic Technology at the Heart of Decision-Making
In Procurement, my area of expertise, and more broadly in any critical decision-making environment, one question remains largely underestimated: Can decision-making be industrialized using non-deterministic technology? Agentic AI is probabilistic, contextual, and not strictly reproducible. But procurement relies on the opposite: traceability, justification, and auditability. And that’s without even considering the human factor. The more effective the AI is, The more it is relied upon, the less it is questioned, and the more vigilance diminishes. Therefore, we can conclude that there is a real risk of a loss of expertise, cognitive dependence, and weakened judgment.
Procurement Decisions and Agent-Based AI: A Major Issue of Accountability
The risk is therefore not merely technological. Because a purchasing decision always entails:
- Financially,
- Contractually,
- Legally
And a decision that can’t be fully explained raises a simple question: Who is responsible? That’s the paradox. The more credible the AI is, the more the risk silently increases. The real danger isn’t that AI makes mistakes. It’s that humans stop verifying. At that point, we should no longer speak of an “augmented buyer” but rather… a “diminished buyer.” So the question isn’t, “How far can AI go?” But how much decision-making should we let it handle? (To be continued) 