Human-audited codes
Gold-standard validation set
MEDICAL HEALTH / AI-ASSISTED CODING
ACE / AUTONOMOUS CODING ENGINEClinical coding.
Evidence behind
every decision.
Reduce the coding burden with EHR-ready AI that proposes codes, surfaces supporting evidence, and directs uncertain cases to human review.
Clinical context to coding clarity
- 01Clinical records
- 02Coding assistance
- 03Specialist review
THE CODING CHALLENGE
Move from clinical notes
to reviewable codes.
ACE analyzes clinical encounters and proposes codes alongside source evidence, confidence scores, and an audit trail. Designed to support efficient, consistent medical coding, it helps healthcare organizations reduce administrative backlogs while maintaining a clear record of how coding decisions were reached.
The solution combines responsible AI, human oversight, and analytics to support transparent, accountable coding workflows across healthcare settings.
ACE BY THE NUMBERS
A foundation for operational validation.
Coding guidelines
Reference knowledge base
Per encounter
Reported demonstration processing time
Confidence range
Reported model confidence, not an accuracy measure
Figures describe the reported validation resources and demonstration results. Encounter time and model confidence do not establish production accuracy or guaranteed performance.
ENCOUNTER EXAMPLE
Motor-vehicle trauma
In the reported demonstration, ACE proposed twelve codes in under thirty seconds: nine ICD-10 diagnoses and three CPT procedures, covering findings from a skull fracture to complex wound repair.
Each code was connected to the supporting words in the encounter note. Conflicts or low-confidence results trigger review by a certified coder.
09ICD-10 diagnoses
03CPT procedures
9 diagnoses · 3 procedures
Source evidence accompanies each proposed code.
KEY STRENGTHS
Built around coding confidence.
Accurate
Ground proposals in coding guidance and evaluate them against human-audited records.
Explainable
Show the supporting encounter text and reasoning context for review.
Scalable
Support increasing encounter volumes through a tiered deployment model.
Secure
Combine isolated infrastructure, encryption, and defined access boundaries.
HOW ACE COMPARES
From coding friction
to a reviewable record.
Three common challenges. Three ways ACE supports a clearer decision.
CHALLENGE 01
Manual coding
Capacity depends on specialist availability.
ACE RESPONSE
Confidence-gated
Score proposals; flag uncertainty for human review.
CHALLENGE 02
Rules-based encoders
Rigid logic can miss clinical context.
ACE RESPONSE
Traceable
Connect each proposed code to source evidence.
CHALLENGE 03
Opaque AI outputs
Missing evidence makes review difficult.
ACE RESPONSE
Auditable
Keep the context behind coding decisions.
THE EVIDENCE PATH
See how a coding proposal becomes reviewable.
- 01
Clinical encounter
Start with the source record.
- 02
Code proposal
Suggest codes with evidence.
- 03
Confidence check
Identify conflicts or uncertainty.
- 04
Reviewable record
Keep evidence and decisions together.
Human review where it matters
Low-confidence or conflicting proposals route to a certified coder for review.
Conceptual workflow · Supports coding review; model confidence is not a measure of accuracy.
THE ACE INNOVATION
Controls built into the coding workflow.
Model risk management
Score outputs and route low-confidence results or conflicting codes to certified human coders.
Verification and validation
Evaluate coding outputs against a gold-standard set of more than 6,400 human-audited codes.
Span-level explainability
Trace each proposed code to the exact passage in the clinical encounter that supports it.
Separation of roles
Keep administrative access separate from the clinical data pathway.
Isolated deployment
Deploy within an isolated virtual private cloud, with encryption at rest and in transit.
EHR-ready exchange
Ingest and export standard encounter formats to support integration with clinical workflows.
BUSINESS MODEL
Managed delivery.
Scale by encounter volume.
ACE’s fully managed SaaS model combines a one-time deployment license with an annual subscription based on encounter volume. Deployment covers assessment, configuration, integration, tuning, and Authority to Operate (ATO) support.
Essential
Up to 1Mencounters / year
Professional
1M–20Mencounters / year
Enterprise
20M–75Mencounters / year
Enterprise Plus
75M+encounters / year
NegotiatedCommercial and government use
Designed for both commercial health organizations and government missions.
Cloud deployment
Infrastructure as code supports cloud portability, including GovCloud. The current implementation runs on Amazon Bedrock with a pluggable architecture.
WHERE ACE FITS
Connected capabilities and markets.
PILOT PARTNERSHIPS