3 numbers I ask for before I touch any AI project

When a team says they want AI, I start with three numbers. If these don't exist, aren't trusted, or nobody owns them, we don't start with AI.

The Three Numbers

1. Time-to-first-response for new leads. How fast do we react, not on average but reliably at p80 during business hours.


2. Conversion lead → meeting or opportunity. Do first touches create real sales motion.


3. Conversion opportunity → closed-won. Can we move a deal through the process predictably.

The Reality in Most Companies

In most companies one of three things is true: the numbers are undefined, the data is dirty, or ownership is fuzzy. That is why so many "AI attempts" disappoint. The tool is fine. The foundation isn't.

My First Step: 1-Day Leak Scan

My first step is a 1-day Leak Scan around those three metrics. We make them real, we clean the data just enough to trust it, and we assign clear ownership so someone actually cares day to day. Only after that do we discuss assistants for reply drafting, enrichment, routing, or scheduling.

Why Order Matters

This order matters. If there is no baseline, every "after" is just a story. If the workflow is broken, AI only speeds up the chaos. But when the manual system holds, you can see whether AI moved the number. That is how you turn "we're doing AI" into "we made more money with the stack we already have."

Common Disappointments

I still meet teams who tried AI first and walked away disappointed. Not because AI is useless, but because they expected it to fix missing ownership, slow response, weak follow-up, and inconsistent handoffs. Fix those, then add AI, and you get lift you can see on the dashboard and in the P&L.

What My Leak Scan Delivers

This is what my 1-day Leak Scan is for. It sets the baseline and the rules so a 10-day audit and a one-month pilot actually land and pay back. If you can name your one metric, we can make it move.