A Foundation Model Built for the Point of Care
Maverick Medical AI announced its Clinical-to-Revenue Foundation Model on September 8, 2026, pitching it as a single model spanning the distance from a physician’s note to a submitted claim. The company, based in Wilmington, Delaware, is framing the release around what CEO Yossi Shahak called a shift “from autonomous coding to Revenue Certainty” — but the more interesting part for CDI teams isn’t the revenue-cycle framing. It’s where the model chooses to intervene.
Most CDI programs are still built around retrospective review: a coder or CDI specialist reads the finished note, spots a documentation gap or a query opportunity, and routes it back to the physician after the fact. That cycle typically runs a day or more behind the encounter, which is long enough that the physician may not remember the clinical detail a query is asking about. Maverick’s architecture tries to close that loop before the note is finalized, which is a meaningfully different workflow than most of what’s shipped under the “AI CDI” label this year.
How CodeAgent Changes the CDI Workflow
From Retrospective Review to Real-Time Queries
The foundation model runs two agents. CodeAgent™ engages physicians while they’re still documenting, flagging missing specificity, coding gaps, and payer-specific compliance risks before the note closes. That’s a point-of-care query model rather than a post-discharge one — the same clinical-query function CDI specialists perform today, but occurring in the encounter instead of days later in the chart.
Physician Burden Versus Documentation Accuracy
The tradeoff CDI leaders will ask about first is interruption. A query that arrives while a physician is mid-note carries a different cost than one that arrives in an inbox the next morning, even if it resolves the ambiguity faster. Maverick’s press materials don’t publish physician-response or query-acceptance rates, and none should be assumed here — that omission is itself worth watching as more of these point-of-care tools reach production. Vendors in this category have generally been quicker to publish coding-accuracy or Direct-to-Bill numbers than physician-experience metrics, and this launch follows that same pattern.
Where mCoder Fits: Autonomous Coding After Documentation Closes
The second agent, mCoder™, picks up once the note is finalized, generating billing-ready codes validated against payer policy and compliance rules. Maverick reports 85%+ Direct-to-Bill performance with full automation coverage in its enterprise production deployments, across specialties including radiology, cardiology, orthopedics, emergency medicine, and pathology. Those figures come from the vendor’s own announcement and haven’t been independently verified; treat them as a starting claim to test in a pilot, not a benchmark.
For CDI programs, the relevant question is less about mCoder’s coding throughput and more about what happens to specialist time once the CodeAgent layer absorbs a share of the queries a human would otherwise generate. That’s a staffing and workflow-design question every program adopting a similar tool will eventually have to answer.
Reading This Against the New AI Query Guidelines
The timing matters. ACDIS and AHIMA’s 2026 query guidelines, which address how AI-generated queries should be reviewed and attributed, arrived just days before Maverick’s launch. A tool that generates physician queries automatically, at the point of care, is exactly the use case those guidelines were written to govern — compliance and CDI leads evaluating a system like this should be checking it against that framework specifically, not treating it as a generic documentation-improvement tool.
What CDI Teams Should Watch
- Query attribution: whether an AI-generated point-of-care query is logged and reviewable the same way a human CDI query is.
- Physician response data: acceptance and override rates for real-time queries, which the vendor hasn’t yet published.
- Specialist role shift: how CDI staffing and review priorities change once first-pass queries are automated.
- Payer-policy validation lag: how quickly mCoder’s rule set updates when payer policy changes mid-quarter.
- Compliance sign-off: whether the model’s coding output is auditable independent of the vendor’s own accuracy claims.
The Bigger Shift
Point-of-care CDI tools are still early, and Maverick is one of several vendors racing to move query generation upstream of the traditional chart review. Whether that improves documentation accuracy or just moves the interruption earlier in the clinical workflow is an open question that this quarter’s pilots — and the compliance frameworks now catching up to them — will start to answer. Health systems that have spent the past two years standing up retrospective CDI programs will need to decide whether a point-of-care layer replaces that work, supplements it, or simply adds a second review step that still has to reconcile against the original query.
For now, the practical move for CDI and compliance leads is the same one that applies to any new coding or documentation vendor: pilot against real chart volume before extending the tool system-wide, and measure query attribution and override rates directly rather than relying on the vendor’s published performance figures.
Programs evaluating tools like this still need a coding layer they can audit and trust regardless of where in the workflow the query originates. Medikode’s automated medical coding platform is built for exactly that kind of transparent, review-ready output.