A recent Forbes piece asked a pointed question: before AI replaces physicians, shouldn't someone ask patients what they actually want? The sharpest answer in it came from ClinicMind's own Chief Medical Officer, Roy Lirov, MD — a physician who also writes software — and it wasn't really about AI. It was about measurement.
"I think we're fooling ourselves by reporting on what's easily measured instead of figuring out how to measure what matters."
— Roy Lirov, MD
His argument: much of the AI-in-medicine conversation grades itself on tasks that are easy to instrument — can a model produce a plausible differential diagnosis? — and treats a good score as progress. The outcomes that actually matter to patients are harder to define and harder to measure, so they quietly fall off the scorecard. What's easy to count crowds out what counts.
We recognize that failure mode — we live in it in revenue cycle
Swap "clinical AI" for "practice financials" and the same trap appears. Practices are handed dashboards full of numbers that are easy to instrument and that climb reliably. The classic one is clean claim rate: the share of claims that clear the clearinghouse's edits on first pass. It is cheap to calculate, it looks great in a monthly report, and it rewards work that has little to do with getting paid.
A high clean claim rate attached to underpriced, underpaid, or written-off claims is a good-looking number sitting on top of a shrinking bank balance. Clean claim rate measures whether a claim was formatted correctly. It says nothing about whether the money arrived.
Measure what pays the practice
The number that matters is net collections per visit — how many dollars actually landed in the practice's account for the average visit. It is harder to instrument because it spans the whole cycle: coding, contracted rates, denials, appeals, patient balances, and write-offs. That difficulty is exactly why it gets skipped, and exactly why it is the one to track.
A few measures pass Roy Lirov's test, because they are tied to dollars and to time-to-dollars rather than to activity:
- Net collections per visit — the dollars you keep, per unit of care delivered.
- Net collection rate — of everything you were contractually owed, how much you actually collected.
- Days in A/R — how long the money takes to arrive, because cash this month is worth more than cash two quarters out.
- Denial recovery — not just how often claims are denied, but how much of the denied dollars you win back.
None of these is as tidy as a clean claim rate. All of them describe whether the practice got paid. That is the point Roy Lirov is making about medicine, and it is the one we have built ClinicMind's revenue cycle service and reporting around: instrument the numbers that pay the practice, not the ones that are easy to report.