Why AI Implementation Fails When Organisations Move Faster Than They Can Govern

This long-form episode of Wicked Problems is a practical conversation about AI adoption after the initial hype. Toby Corballis and Marcellus Lindsay look beyond the promise of automation and ask what happens when AI is introduced into organisations that cannot yet govern its cost, risk, data, accountability or human impact.

The discussion moves between real-world implementation, change leadership, AI literacy, trust, hallucinations, token costs, agentic AI, and the familiar pattern of technology running ahead of organisational capability. The point is not that leaders should avoid AI. It is that they need enough understanding to ask better questions before they scale it.

Marcellus brings the perspective of a consultant implementer: someone who can work with senior leaders while also understanding what happens on the ground. Together, he and Toby argue for controlled speed, domain expertise, human-in-the-loop review, and governance that helps organisations move quickly without building on flawed foundations.

Key themes
- AI governance, risk and accountability as leadership responsibilities, not technical afterthoughts.
- Why many corporate AI initiatives fail to produce clear return on investment.
- The gap between exponential technology change and slower organisational change.
- How AI hallucinations, confidence and error rates create hidden implementation risk.
- Why human trust in AI can exceed human trust in other people.
- Token pricing, cost controls and the danger of unbounded AI usage at scale.
- The role of domain expertise in judging where AI should and should not be applied.
- Why agentic AI needs stronger guardrails than simple chat-based experimentation.
- How flawed assumptions early in a programme compound into expensive failure later.
- Why AI literacy should be basic, practical and principle-based for senior leaders.

To register intrerest in one of the forthcoming webinars mentioned, sign up at Strategic Flow