IN THIS ISSUE The receipt | The list, before and after | What moved each row | Why it mattered |
In Tuesday's piece I made a claim. AI is the tool, judgment is the product, and what a buyer pays for is the judgment loop wrapped around the model. That is easy to assert. It is harder to show. So here is the receipt.
A few weeks ago I delivered a market map for a PE-backed pain management platform. I have shared pieces of this engagement before. This time I want to put the actual before-and-after in front of you, because it is the cleanest example I have of what the judgment layer does that the model cannot.
They scoped me to a tight buy box: independent, physician-owned interventional pain practices across a handful of metro markets in a single Southeast state. Inside that footprint they were already tracking a list. Good team, real process, a live pipeline in a deal tracker. What they hired me to answer was not "find me targets." It was closer to "tell me what we are missing, and tell me who to call first."
The List, Before and After
Inside the buy box, the map held 20 practices. They were already tracking 14 of them. Six we surfaced: five they had never seen, and one they had written off. Nine we ranked Tier 1.
Same buy box, 20 practices. They were tracking 14; we surfaced 6 (five new, one recovered); nine ranked Tier 1.
The six net additions are not the headline. The headline is what ordered the whole list. Every practice carries a transaction-readiness signal, a score that estimates how close the owner actually is to a deal. The inputs are roughly what you would expect: roster movement, physician departures, lease and property timing, reimbursement exposure, local competitive pressure. The categories are not the secret. The work is pulling each one out of scattered public records and scoring it into a single read, High to Low. That score, not gut feel, is what flips a name from "on the list" to "call this quarter."
The model assembled the raw material. The judgment decided what it meant.
What Actually Moved Each Row
Look at what happened in each of those rows. The AI pulled the taxonomy code, the license record, the patient reviews, the county filing, the archived web page. Fast, and genuinely useful. But in every single row, the thing that moved the call list was a judgment: that a miscoded practice was worth re-verifying instead of trusting the label, that a stale website was an operational signal and not noise, that a five-physician roster was really two, that a competitor sitting three miles away changes your sequencing.
None of that came off the shelf.
KEY TAKEAWAY The model gets you to the evidence. The judgment tells you what to do about it. That is the part you are paying for, and it does not get cheaper when everyone has the same models. |
One aside, because people assume the opposite: almost all of the timing inputs here are public. Licensing, the national provider registry, county and property filings, archived sites. They are just scattered and uneven, so nobody assembles them into one ranked call list. The assembly is the product, and the judgment about what the assembled picture means is the moat.
Why It Mattered
The VP of M&A brought us in on this one. The report went up to their CEO, who turned around and extended us into additional markets. That is the part that counts. Not that the map was thorough, but that it changed what the team did next: who they called first, which name they recovered, which number they did not overpay.
The case I made on Tuesday holds: if you are scoping an AI-enabled engagement, ask to see a redacted sample before you scope anything, because the judgment loop shows up in the artifact, not the proposal. The crosswalk above is that sample. If your reaction reading it was "I could not pull that out of a database," that is exactly the point.
AI did not build this map. I built it, faster than I could have three years ago, using AI at every step where assembly was the task. The judgment was mine. In a market where everyone now has the same models, that was always going to be the only thing separating one target list from another.
Tuesday was the claim. This is the receipt.
If you are running corp dev or BD inside a healthcare platform and you want to see what this looks like against your own pipeline, reply to this email. I will show you a redacted version of the kind of map we build.
-Shawn
If you made it this far, one quick ask. I write these long because I'd rather show the work than tease it. But I want your honest read on the length. Tap below, and add anything else in the box. I read every one.
Honest answer: how's the length?
This newsletter is for informational purposes only and does not constitute investment, legal, or financial advice.