Necessary, and Sufficient.

Most conversations about AI adoption start in the wrong place — with the technology: which model, which platform, which use case to pilot first. This paper argues that five leadership conditions have to be in place before any of that matters, and that getting four right while missing the fifth is the most common way organisations spend real money on AI and have little to show for it.

The first four conditions are necessary. None of them, even together, are sufficient. This paper walks through each in turn, then makes the case for why the fifth — defining what a good outcome actually looks like, and measuring honestly against it — is what turns four correctly built capabilities into an actual roadmap.

Key takeaways

• Ownership belongs at the top. AI is a business transformation, not a software project, and it cannot be delegated to the technology function.
• Data comes first. Every AI decision rests on data that needs named business owners, and governance you could show a regulator.
• Every function faces the same problem. The CFO, CRO and CHRO each find their operating cadence broken by AI — getting ahead of it is the CEO's job; checking that someone sees it whole is the Board's.
• The business has to be agile, not just the technology. Faster delivery only helps if the organisation can absorb and direct it.
• Outcomes turn four good conditions into a direction. Necessary is not the same as sufficient — the roadmap at the end of the paper lets your team score itself.