When AI Codes the Note: OpenEvidence's Point-of-Care Coding Intelligence

AI medical coding automation has taken a new direction. On March 24, 2026, OpenEvidence launched Coding Intelligence, a feature that applies automatic ICD-10 and CPT code suggestions directly from clinical documentation — at the moment a physician finishes a note. Rather than routing encounter data through a separate billing workflow, the system surfaces codes, E/M level recommendations, and Medical Decision Making (MDM) rationale before the patient leaves. It is a meaningful shift in where AI medical coding automation happens: upstream, inside the clinical encounter, not downstream in a billing queue.

What Is Coding Intelligence and How Does It Work?

Coding Intelligence is built into OpenEvidence's clinical documentation platform, Visits. When a physician completes a note, the system automatically generates three outputs: ICD-10 diagnosis codes derived from the documented conditions, E/M level recommendations with a full MDM breakdown written directly into the note, and CPT procedure code suggestions based on what occurred during the visit.

The MDM component is particularly notable. Rather than leaving documentation of medical complexity to the physician after the fact, the system generates the reasoning automatically — including whether the visit qualifies as billing by complexity or by time, and the supporting rationale embedded in the note. CPT suggestions also include expected RVU values alongside each code, so physicians can sequence codes correctly before the claim goes out.

Why This Approach Is Different

Most AI coding tools work after the clinical encounter. A coder receives a completed note, runs it through a coding engine, reviews suggestions, and routes corrections back through a billing queue. The cycle can take days. OpenEvidence flips this by embedding code generation into the documentation moment itself. When the physician finalizes the note, the codes are already drafted. The bottleneck shifts from billing operations to clinical documentation quality — which is where it should be.

The integration also changes who owns coding accuracy. In a traditional back-end workflow, coders carry the burden of interpreting incomplete documentation. With point-of-care coding, the physician sees the suggested codes before signing the note — and can correct them, add missing diagnoses, or clarify the MDM rationale in real time. This creates a feedback loop that improves documentation quality over time rather than just patching billing errors after the fact.

What This Means for Medical Coders and CDI Specialists

For medical coders, point-of-care AI coding is not a replacement — it is a quality filter that arrives earlier in the workflow. Rather than building codes from scratch on incomplete notes, coders review and validate AI-generated suggestions against documented evidence. The role shifts toward auditing, exception management, and clinical documentation improvement (CDI) consultation rather than

• Audit trail transparency: Does the AI document which clinical evidence supported each code suggestion? Auditors will ask.

• Physician override rates: High rejection rates in specific service lines signal model drift or documentation style mismatches worth investigating.

• Payer scrutiny of AI-assisted claims: CMS has not issued explicit guidance on AI-generated MDM documentation. Coders should confirm the rationale reflects actual clinical decision-making, not just AI-drafted text.

• Integration with back-end RCM workflows: Point-of-care suggestions need to connect cleanly with downstream systems to avoid duplicate or conflicting code sets.

The Bigger Picture: AI Coding Is Moving Upstream

OpenEvidence Coding Intelligence is one data point in a broader trend. Corti Symphony from April 2026 uses a multi-agent architecture to code from clinical notes. Amazon Connect Health surfaces ICD-10 and CPT codes from patient conversations. AI medical

This matters for how coding teams are structured, CDI workflows are designed, and what skills coders need going forward. The shift is not hypothetical — it is shipping now.

If your organization is evaluating AI medical coding automation — whether at the point of care or in your RCM back-end — see how Medikode's automated medical coding platform integrates with your existing workflows.