Part 1: The AI Executive

Part 2: The not so Sexy AI & Data Governance

Part 3: Measure What Matters

Part 4: A new software Play Book

Part 5: The Engine Room

After wading through the quagmire of enterprise governance, you inevitably hit the next corporate obsession: measuring return on investment.Lately, boardroom decks and conference stages are flooded with rants about companies seeing zero ROI from their artificial intelligence bets. That zero is often real, but treating it like an indictment of the technology completely misses where we are. Most companies are only in Year 1 or 2 of an emerging paradigm shift. Applying mature, rigid KPI frameworks right now makes no sense. The last platform shift this significant was the commercial internet, and before that technology reshaped the global economy, it triggered 1 of the largest market crashes in modern history. In these early stages, you cannot treat discovery like a solved equation. You have to stick your finger in the air and see which way the wind is blowing. Your metrics can and must evolve alongside your capability.

I have a few simple themse I follow along the journey to commercial AI :

1. Measure Adoption: Drive adoption above all else.
2. Measure Behaviour: Steer the Engineering Practices
3. Measure Value: Derive the meaningful outcomes

If people are not using the tools, there is nothing to govern, nothing to measure, and zero chance of creating value. Yes, teams will overspend during the discovery phase. Early on, when Claude first hit the scene, I took a sudden $6000 surprise bill on the chin. It stung, but it was tuition. Today, I know how to start small and manage costs early. Once the novelty fades, teams naturally settle into sustainable, predictable usage habits. Only then can you graduate to measuring behavior. How are engineers using models to improve software craft? Are they leveraging AI to scaffold and planning strategies, or are they taking it too far and generating bloated, Markdown repos? What does their ratio of clean code generation look like compared to living documentation? From there, you can finally move to hard value metrics. Even those should start simple: How many manual workflows did we eliminate? How much did we compress cycle times? Did that new customer feature reach production weeks ahead of schedule? Those results do not come from premature accounting spreadsheets. They come from letting the team learn the tools 1st. Instead, I constantly sit in KPI deliberation sessions listening to leaders agonize over how employees might game the system. I understand caution in a large enterprise, but we need to extend some basic trust to the adults we hired. If you treat your people like potential fraudsters before they even adopt the tool, you kill experimentation in the crib.

Measure adoption 1st, study engineering behavior 2nd, and let real value metrics emerge as the team finds its feet. If you demand a precise financial return before the dust settles, you will starve your best initiatives before they ever have a chance to breathe. Yet even with sensible metrics and pragmatic governance, tools and numbers alone will not save you. In Part 4, we will tackle the real engine of transformation: why your engineering culture itself must radically evolve if enterprise AI is ever going to succeed