Let the business outcome lead.

Revenue, costs, innovation, and risk give the work a purpose. I define using AI well by its contribution to those outcomes. Speed and proficiency matter through what they make possible.

In practice

The first thing I ask a team is which line of the P&L a plan is supposed to move, and how they’d know. Most plans can’t answer. They name a tool and a team, and the outcome is implied. So I write the expected path to value in plain language before anything gets built: less time preparing for customer conversations, more useful conversations, faster learning about what matters, more business. Each link in that chain is a hypothesis, and each one gets inspected.

The tradeoff

Outcome-first can starve exploration. Some of the most valuable things I’ve built started as a curiosity with no business case, and a strict outcome test would have killed them at the door. So I keep two lanes: work that has to earn its place against an outcome, and a small, bounded budget for experiments whose only job is to teach me something.

Go deeper

The field guide starts here: Find the constraint worth changing.