AI leadership handbook for Heads of AI, CTOs and VPs
Most companies don't have an AI technology problem. They've bought tools, run pilots and hired people, and still nothing is in production that changes a number the business tracks. These pages are for the people who own that gap, Heads of AI, CTOs, VPs of engineering and the executives they report to, and they're organised around the decisions you'll make in roughly the order you'll face them.
Judgment is deciding what not to build
As a leader in AI, the two scarce resources are your team's time and your budget. Most of the value of this handbook is in deciding what to stop, so the few projects that matter get both.
Decisions, in order
| Stage | Decision | Page |
|---|---|---|
| Diagnose | Why isn't our AI spend producing results? | Why AI projects stall |
| Diagnose | Are we ready, and where are we weakest? | AI readiness assessment |
| Decide | What's our strategy, on one page? | AI strategy on one page |
| Decide | Which use cases first, and which to kill? | First three use cases |
| Decide | Hosted APIs or our own models? | Cloud APIs vs open-source LLMs |
| Decide | Build, buy or partner? | Build vs buy vs partner |
| Decide | Is it worth the money? | ROI and cost estimation |
| Deliver | How do pilots reach production? | POC to production |
| Deliver | Who do we need? | Structuring an AI team |
| Deliver | How do we hire? | Hiring AI engineers |
| Deliver | How do we train the people we have? | Upskilling engineers |
| Control | How do we govern without slowing down? | AI governance |
| Control | How do we judge vendors? | Vendor evaluation scorecard |
| Report | Is it being used, and does it help? | Measuring AI adoption |
| Report | What does the board need to hear? | Board AI reporting |
| Start | I'm new in the role. Where do I start? | Head of AI first 90 days |
How these pages differ from the engineering docs
They don't ask you to write code. They do link to the engineering pages where a decision depends on a technical fact, like what drives inference cost or why an agent needs approval steps, because a leader who understands the trade-offs makes better calls and gets better estimates from the team.
From the floor
In the founding session of the AI leadership cohort, I said the two most important resources in the age of AI are your time and your budget, and that judgment comes from dropping ideas: knowing which projects not to work on, even when you've spent six months on one and your ego is attached to it.