Procode, an AI-powered RCM company for private practice surgeons, has closed a $10 million Series A led by Health Velocity Capital. The round funds additional surgical billing acquisitions, follows the acquisition of The Auctus Group, and comes alongside the publication of the first peer-reviewed study validating that Procode's AI outperforms both professional coders and every major general-purpose LLM on complex surgical coding.
The numbers in the research paper are striking enough to warrant a close read: 86.7% overall accuracy on CPT coding versus 42.5% for human auditors and 35.8% for OpenAI GPT-5 — the best-performing general-purpose LLM in the study. On complex cases with four or more CPT codes, Procode held at 80% while GPT-5, Gemini 2.5 Pro, and Claude Sonnet 4.5 ranged from 5% to 12.5%.
Procode AI's overall CPT coding accuracy vs. 42.5% for human auditors and 35.8% for OpenAI GPT-5 — the next-best performer — in a peer-reviewed study published in Plastic & Reconstructive Surgery Global Open.
Most funding announcements in RCM AI come with performance claims that can't be independently verified. Procode took a different approach: they submitted their results to a peer-reviewed journal before the round closed.
The study, "Artificial Intelligence for Automated CPT Coding in Plastic and Reconstructive Surgery Using a Fine-Tuned Hybrid LLM," was published in Plastic & Reconstructive Surgery Global Open — the official open-access journal of the American Society of Plastic Surgeons. It reviewed 120 operative reports across three difficulty levels (40 easy, 40 medium, 40 hard) and compared Procode's fine-tuned hybrid LLM against GPT-5, Gemini 2.5 Pro, Claude Sonnet 4.5, and external professional auditors.
| Model / Coder | Overall Accuracy | Hard Cases (4+ CPT codes) |
|---|---|---|
| Procode AI (fine-tuned) | 86.7% | 80.0% |
| Human Professional Auditors | 42.5% | — |
| OpenAI GPT-5 | 35.8% | ~12.5% |
| Google Gemini 2.5 Pro | — | ~10% |
| Anthropic Claude Sonnet 4.5 | — | ~5% |
The performance gap on hard cases is the most important number here. Complex surgical procedures — the ones with multiple procedures, modifiers, and documentation requirements — are exactly where coding errors cost the most. Undercoding loses revenue. Overcoding creates audit exposure. Inconsistency across coding staff creates both. A model that holds 80% accuracy on the hardest cases in its domain while general-purpose LLMs are at single digits is genuinely different in kind, not just degree.
The gap between Procode and GPT-5 on hard cases (80% vs. ~12.5%) isn't a training data problem — it's a domain-specificity problem. Surgical CPT coding requires operative report interpretation, knowledge of bundling rules, modifier hierarchies, and payer-specific adjudication patterns. General-purpose models weren't fine-tuned on this. Procode was. That's the entire thesis.
The $10M round funds two additional acquisitions to be announced in the coming months. Procode's playbook: acquire the best surgical billing companies, layer in their AI, and drive measurably better outcomes. The current portfolio:
The Auctus Group acquisition — the leading RCM company for plastic surgeons and dermatologists — was the first execution of this playbook. CEO Jeff Cripe put a specific number on it: "We're on track to double The Auctus Group's revenue and quintuple its EBITDA margin." That level of operational specificity in a funding announcement is either a confident benchmark or a liability — and either way it signals the team is tracking it closely.
Worth noting: Procode bills on contingency. They get paid when clients get paid. That alignment is structurally different from software licenses collected regardless of outcomes — and it means margin improvement is a direct function of AI accuracy, not just customer count.
Health Velocity Capital doesn't need an introduction in healthcare RCM circles — their portfolio includes Change Healthcare, Teladoc, Livongo, and a long list of companies that built or defined markets. When a firm with that track record leads a Series A in surgical billing AI, the thesis is worth unpacking.
Grant Blevins, Partner at Health Velocity Capital: "We have grown increasingly convinced that this market is ready for a company that can finally deliver true end-to-end automation. Procode is uniquely positioned to do that by giving small billing practice owners a best-in-class AI strategy no traditional buyer can match, while preserving the specialty expertise that made their businesses valuable."
The framing matters. "No traditional buyer can match" is a direct shot at the PE consolidators who have been rolling up surgical billing practices for the past five years. The traditional acquirer extracts EBITDA. The Procode model claims to expand it — by adding AI to operations that couldn't afford to build it themselves.
Private practice surgeons represent a segment that has been almost entirely ignored by the enterprise RCM AI wave. The Commures, Candid Healths, and R1s of the world are built for health systems and large medical groups. The unit economics don't pencil for individual surgical practices. The contract sizes are too small, the fragmentation is too high, the customization required per specialty is too deep.
Procode's answer is the roll-up model: acquire specialty-specific billing expertise, apply AI at the workflow level, and amortize the tech cost across a growing portfolio of practices. If the AI accuracy holds and the acquisition integration doesn't dilute outcomes, this is a real wedge into the 350+ practices they've already acquired through Auctus alone.
If you're currently using in-house coders or a generic billing company for surgical coding, this research is worth reading directly. The peer-reviewed accuracy data on complex cases represents a real benchmark. The contingency billing model aligns incentives in a way that per-encounter or per-seat software doesn't. The risk: you're an early customer of an AI platform that hasn't yet proven scale across multiple surgical specialties outside of plastics and derm.
Procode has one of the more credible origin stories in RCM AI — peer-reviewed benchmarks, surgical specialty focus, and a contingency billing model that forces accountability on outcomes. The next 18 months will determine whether this is a genuine platform or a single-specialty niche:
Procode's Series A is notable for one reason that sets it apart from most RCM AI raises: they brought peer-reviewed evidence to the announcement. In a space where claims about AI accuracy are routine and verification is nearly impossible, a published comparative study with specific benchmark numbers is a different category of proof.
The thesis — acquire surgical billing expertise, embed domain-specific AI, bill on contingency — is structurally sound for private practice. The risk is execution at scale and generalization across specialties. But the foundation is real, the investor knows healthcare deeply, and the 86.7% accuracy claim is now on record in a peer-reviewed journal.
That's a harder thing to walk back than a slide deck.
Daily coverage of payor policy, denials, and vendor moves — written for people who live in revenue cycle.