10decodersEvery code, traced back to the clinical text.
Coders at a regional health network read every page of every clinical document by hand to find diagnosis and procedure codes, with no way to show which sentence produced which code. 10decoders deployed MedCode AI — it reads the document, extracts ICD-10 and CPT codes in a single pass, and links each one back to the passage it came from.
Coding is a reading problem, not a lookup problem.
The network's revenue cycle team worked through clinical encounter documents page by page — physician notes, lab reports, discharge summaries — pulling out the diagnosis and procedure references and matching each to a code. The reading was the slow part, and nothing in the process recorded why a given code had been assigned. When a payer queried a claim, answering meant opening the document again and finding the sentence from memory. Managers had no view of which documents were done, which were pending, and which had been flagged.
10decoders built MedCode AI to do the reading and keep the receipts. A clinical document is uploaded, the platform reads it end to end, identifies every diagnosis and procedure reference, and maps each to the correct ICD-10 and CPT code with a confidence score. Every code is anchored to its source passage in the original file, so verification is a click rather than a re-read. The batch dashboard shows the whole queue — status, code counts, confidence flags and assignments — and coders accept, override or export from the review interface.
From page-by-page reading to traceable extraction.
The challenge
- Coders read every page of every clinical document by hand to find codeable references
- No traceability — nothing recorded which passage had produced a given ICD-10 or CPT code
- Answering a payer query meant reopening the document and locating the evidence from memory
- No way to confirm that every codeable reference in a document had actually been found
- No batch view of which documents were coded, pending or flagged across the team
Our approach
- NLP reads the full document and identifies diagnosis and procedure references in context
- ICD-10 and CPT codes extracted in a single pass, each with a confidence score
- Every code anchored to its source passage, highlighted in the original document
- Payer-specific rule validation on procedure codes before the set is exported
- A batch dashboard giving managers status, code counts and review flags across the queue
Built for the full coding workflow.
AI sits at the extraction layer, where the reading burden is. Coders stay in review, and every decision keeps its evidence.
Diagnosis code extraction
NLP reads clinical text and maps every diagnosis term to the correct ICD-10 code, with a confidence score and a source link on each result.
Procedure code mapping
Procedure references across physician notes, operative reports and fee sheets are mapped to CPT codes and validated against payer-specific rules.
Source traceability
Each extracted code links to the exact sentence that produced it. Verification and audit response become a click instead of a re-read.
Batch queue visibility
Managers see every document in the processing batch — status, code count, confidence flags and review assignments — without chasing individual coders.
Review & export
Coders accept or override each extracted code and export the verified set in standard formats, ready for claim submission.
Secure document handling
Documents are processed in a HIPAA-compliant environment, encrypted at rest and in transit, with role-based access and a log of every coder action.
Four capabilities. One platform.
MedCode AI covers the path from document ingestion through to a verified, export-ready code set.
ICD-10 extraction engine
NLP-powered diagnosis code extraction from any clinical document, with a confidence score and source passage traceability on every code returned.
CPT procedure code engine
Procedure code mapping across physician notes, operative reports and fee sheets, with payer-specific rule validation and full coder override support.
Batch processing dashboard
Real-time visibility across every document in the batch — status tracking, extracted code counts, confidence flags and team-level coder assignment management.
HIPAA-compliant architecture
Secure document handling with encryption at rest and in transit, role-based access control, and a complete audit log of every coder action and code export.
From document upload to a verified code set.
Four steps take a raw clinical document through to a reviewed, export-ready result.
Upload documents
Encounter PDFs are uploaded individually or in batch and queued for processing — no reformatting or manual prep.
AI reads and extracts
The platform reads the full document, identifies every diagnosis and procedure reference, and maps each to a code.
Verify against source
Coders check each code against the highlighted passage that produced it, and accept or override with the evidence in view.
Export for submission
The verified code set exports in standard formats, ready for claim submission, with every action logged.
What each team actually sees.
Role-scoped views and permissions — coders work the document, while compliance and revenue cycle watch the batch and the log.
Verify instead of search
- Upload a clinical document and receive its ICD-10 and CPT codes without reading page by page
- See the source passage for every code, highlighted in the original, before accepting or overriding
- Export verified code sets straight from the review interface in claim-ready formats
The full batch and the full log
- Document status, code counts, confidence flags and coder assignments across the whole team workload
- Every extraction, override and export timestamped — a complete trail for payer queries and reviews
- Confidence thresholds and payer rule flags surfaced before a claim is submitted, not after
Defensible codes, in your control.
What changed for each group — and the controls the platform is built on.
Reviewing, not hunting
The reading pass moves to the platform. Coders open a document with its codes already extracted and anchored, and spend their attention on the judgement calls rather than the search.
Evidence attached to every code
Each ICD-10 and CPT code carries a link to the passage that produced it. An audit response is a lookup rather than a document re-review.
HIPAA-aligned by design
Documents are processed in a HIPAA-compliant environment with encryption in transit and at rest, role-based access and NDA-bound teams. Source code and IP ownership remain with the client.
