Where the Rules Run Out
Doximity’s survey found 28% of physicians have clear AI policies and approvals at work. The gap by setting is the story: 53% in government hospitals and clinics, 14% in private practice. Here is every state’s physician workforce from CMS’s clinician file, split by the size of the practice each doctor bills through. Slide the survey’s rates and the cutoff for “small,” and see where the written rule runs out.
Read the Doximity report →
Or read today’s newsletter →
Primary source: Doximity, “2026 Doctor-Driven AI Report” (3,379 U.S. physicians, 20 specialties, surveyed September 2026; not population-based), and the
Nov. 12 release. Data via MIMI Labs: CMS Doctors and Clinicians National Downloadable File, October 1, 2026 snapshot (MD and DO credentials only).
The survey didn’t break out practices by size, and CMS doesn’t record whether a practice is “private.” What CMS does record is how many clinicians bill under each group. Small groups are the closest public proxy for private practice, so that’s the lever here. Each physician counts once, in the largest group they bill through.
Physicians shown
In “small” practices
Projected with clear rules
Projected without
Each bar is 100% of a state’s (or specialty’s) Medicare-enrolled MDs and DOs, split by the number of clinicians in the largest group they bill through. Red bands are the groups you’ve called “small.” The number at the right is the projected share working under clear AI policies: small × your small-practice rate + the rest × the larger-group rate. With the anchor on, the larger-group rate is solved so the 50 states and DC blend to 28%. Hover a bar for counts.
80/20 · what the bars say
CMS lists MDs and DOs across the 50 states and DC. bill through groups of fewer than 10 clinicians, and through groups of 500 or more. With small practices at the survey’s 14% and the rest anchored so the country blends to 28%, the larger groups sit near . The spread between states is narrower than you’d guess, projected under clear rules, because most doctors everywhere already bill through big groups. Your state isn’t the variable. Your employer is. Puerto Rico, left off the chart, is the outlier: of its physicians bill solo. Switch to specialties and the spread widens: work in small groups, against .
Read this before quoting a number
- The projection is arithmetic, not a finding. Only the 28%, 14% and 53% are from the survey. Doximity didn’t report rates by group size, state or specialty, or how many respondents sat in each setting. A per-state number here is the survey’s national rate re-weighted by local practice mix, nothing more.
- The best-ruled setting is mostly missing. Government hospitals and clinics reported 53%. VA and military physicians largely don’t bill Medicare, so they’re thin in this file. That’s why the 53% is a slider, not the default for large groups.
- Group size isn’t ownership. A 600-doctor independent multispecialty group counts as “large” here, and a three-doctor clinic owned by a health system counts as “small.” CMS reassignment data can’t tell them apart.
- Who’s counted. Medicare-enrolled clinicians with an MD or DO credential in the file (about 35,000 records with a blank credential are left out). Pediatricians and psychiatrists who don’t bill Medicare are undercounted; Puerto Rico and other territories are excluded from the chart.
- The survey sells something. Doximity markets AI tools to the doctors it surveyed. That matters more for its adoption numbers than for the 28%, which flatters nobody.