📅 Book a 30-min Demo📞 Call/text (571) 293-0242
Healthcare · AI Course · MED-3

Medical Coding with AI: ICD-10, CPT, and Denial Prevention

AI-assisted coding that improves accuracy rather than just speed — code suggestion, documentation gap detection, denial prevention, and staying clear of upcoding.

Last updated:

The Short Answer

AI coding assistance improves specificity and prevents denials, but suggesting a higher-paying code the documentation does not support is upcoding. ibl.ai runs coding assistance inside the organization where you own all the code and the data, with a compliance guardrail and an audit log for every suggestion.

On ibl.ai you own all the code and the data, run it model-agnostic across any LLM, and pay with no per-seat pricing — so you can deploy anywhere, from your own cloud to a fully air-gapped network.

The full course design is published below — every module, its objectives and hands-on activity, the capstone, and every source it cites.

What is this course about?

Coding is rule-governed, high-volume, and directly tied to revenue, which makes it an obvious AI target and a compliance risk. This course covers code suggestion from documentation, the specificity gaps that drive denials, documentation queries to clinicians, and the compliance line an assistant must never cross toward upcoding.

Who is this course for?

  • Medical coders and coding managers
  • Clinical documentation improvement specialists
  • Revenue cycle leadership
  • Compliance staff covering coding

What do I need before starting?

  • Medical coding experience or CDI background
  • Familiarity with your coding workflow

What will I be able to do afterwards?

  • Identify where coding errors actually originate
  • Use AI code suggestion with appropriate coder review
  • Detect the specificity gaps that drive denials
  • Generate clinician documentation queries that get answered
  • Enforce a compliance guardrail preventing upcoding

What does each module cover?

1

Where do coding errors actually originate?

45 min

Tracing errors to their source, which is usually documentation rather than coding.

Objectives

  • Trace coding errors to their origin
  • Distinguish documentation from coding errors
  • Target intervention where errors originate

Topics

Error origin analysisDocumentation versus codingDenial root causesIntervention targeting

Activity. Analyze a denial sample and trace each to its origin.

2

How does code suggestion work and where does it fail?

50 min

Suggestion accuracy by code family and the categories that need the most review.

Objectives

  • Measure suggestion accuracy by code family
  • Identify systematically error-prone categories
  • Target coder review accordingly

Topics

Suggestion accuracyCode family variationError-prone categoriesReview targeting

Activity. Measure suggestion accuracy across code families on real historical charts.

3

Why does specificity drive denials?

45 min

Unspecified codes and the documentation gaps behind them.

Objectives

  • Identify unspecified code usage patterns
  • Trace them to documentation gaps
  • Quantify the denial cost

Topics

Specificity requirementsUnspecified code usageDocumentation gapsDenial cost

Activity. Quantify unspecified code usage and its denial cost in one service line.

4

How do you write a query a clinician answers?

50 min

Documentation queries that are compliant, non-leading, and actually get responses.

Objectives

  • Write compliant non-leading queries
  • Design queries clinicians will answer
  • Track query response rates

Topics

Query complianceNon-leading constructionResponse ratesClinician burden

Activity. Write ten queries and have a clinician rate whether they would answer each.

5

Where is the line before upcoding?

50 min

The compliance boundary between improving specificity and suggesting unsupported codes.

Objectives

  • Define the compliance boundary precisely
  • Implement it as a structural guardrail
  • Audit for boundary violations

Topics

Upcoding definitionSpecificity versus inflationStructural guardrailsAudit

Activity. Implement the guardrail and attempt to get an unsupported higher-paying code suggested.

6

How do you analyze denial patterns?

45 min

Denial analysis that produces prevention rather than just appeals.

Objectives

  • Analyze denial patterns by payer and code
  • Identify preventable denial categories
  • Close the loop into documentation improvement

Topics

Denial pattern analysisPayer variationPreventable categoriesLoop closure

Activity. Analyze a denial set and identify the top three preventable patterns.

7

How do you measure coder productivity honestly?

40 min

Productivity measurement that does not degrade accuracy in pursuit of speed.

Objectives

  • Measure productivity without degrading accuracy
  • Track accuracy alongside throughput
  • Detect speed-accuracy trade-offs

Topics

Productivity measurementAccuracy trackingTrade-off detectionIncentive design

Activity. Design a measurement approach that tracks both and would detect a trade-off.

8

Building the coding assistant

50 min

The lab module: an assistant with a compliance guardrail and complete audit logging.

Objectives

  • Build the assistant with a compliance guardrail
  • Log every suggestion and coder decision
  • Validate against historical coded charts

Topics

Assistant buildGuardrail implementationAudit loggingHistorical validation

Activity. Validate the assistant against 50 previously coded charts and measure agreement.

What is the capstone project?

Coding assistant with a compliance guardrail

Build a coding assistance workflow with accuracy measured by code family, specificity gap detection, compliant documentation queries, a structural upcoding guardrail, denial pattern analysis feeding documentation improvement, and full audit logging.

Deliverable: A validated coding assistant with guardrail testing and historical agreement measurements.

How are learners assessed?

  • Guardrail tested — an unsupported higher-paying code must not be suggestible
  • Agreement measured against 50 previously coded charts
  • Queries rated by a clinician for whether they would answer

What ships with the course?

Facilitator guide

Session-by-session running order, discussion prompts, and the questions that reliably derail a room.

Learner workbook

Exercises, checklists, and the templates each module's activity produces.

Hands-on lab environment

A sandboxed ibl.ai deployment so exercises run against real agents, not screenshots.

Assessment bank

Scenario questions and rubric criteria mapped to each stated learning outcome.

Source bibliography

Every primary regulation and standard cited on this page, linked and dated.

Which AI agents does this course use?

The hands-on modules run against agents already deployable on the ibl.ai platform for healthcare.

Where does the course material come from?

Every module is grounded in primary sources — the regulation, standard, or research itself, not a summary of it. Each was resolved at authoring time.

  • ICD-10-CM

    Centers for Disease Control and Prevention

    The diagnosis coding system and its official guidelines.

  • Centers for Medicare and Medicaid Services

    CMS

    Coding, billing, and documentation requirements.

  • HIPAA

    U.S. Department of Health and Human Services

    Privacy requirements for the charts the assistant reads.

  • HealthIT.gov

    ASTP/ONC

    EHR integration guidance for the coding workflow.

Delivery notes

Binding guidance for anyone preparing and delivering this course.

  • Module 5's guardrail must be structural and tested adversarially. Upcoding via an AI assistant is a False Claims Act exposure and a soft warning will not prevent it.
  • Module 4's non-leading query construction is a compliance requirement, not a style preference. Have a compliance officer review the examples.
  • Module 7's speed-accuracy trade-off is the risk of deploying coding AI under productivity pressure. Design the measurement to detect it before it becomes an audit finding.
  • Use synthetic charts with known correct codes. Agreement measurement needs ground truth and real charts cannot be used in a workshop.
  • Coordinate with MED-2 — documentation quality is the upstream cause of most coding problems and the two courses should share the gap analysis.

Why run AI training on a platform you own?

You own the course, not a licence to it

Course content, learner data, and the platform run inside your perimeter — you own all the code and the data.

Model-agnostic delivery

Run the course's AI components on any LLM — Claude, GPT, Llama, Gemini, Command — and switch anytime.

No per-seat training licences

Usage-based or self-hosted, so cost tracks actual use rather than headcount.

Deploy anywhere

Cloud, private VPC, on-premise, or fully air-gapped — including for cohorts that cannot use public AI tools.

Frequently asked questions

What does the Medical Coding with AI: ICD-10, CPT, and Denial Prevention course cover?

Coding is rule-governed, high-volume, and directly tied to revenue, which makes it an obvious AI target and a compliance risk. This course covers code suggestion from documentation, the specificity gaps that drive denials, documentation queries to clinicians, and the compliance line an assistant must never cross toward upcoding. It runs 5.5 hours across 8 modules across 8 modules, at intermediate level, and closes with a capstone: Coding assistant with a compliance guardrail.

Who should take Medical Coding with AI: ICD-10, CPT, and Denial Prevention?

It is written for Medical coders and coding managers, Clinical documentation improvement specialists, Revenue cycle leadership, Compliance staff covering coding. Prerequisites: Medical coding experience or CDI background; Familiarity with your coding workflow.

Can we run this course on our own infrastructure?

Yes. ibl.ai is model-agnostic and deploy-anywhere — cloud, private VPC, on-premise, or fully air-gapped — and you own all the code and the data. Cohort data, submissions, and any material learners upload stay inside your perimeter, which matters for healthcare teams that cannot send work to a public AI tool.

How do we get access to Medical Coding with AI: ICD-10, CPT, and Denial Prevention?

Request access and we will set it up for your cohort — hosted by ibl.ai, or running against your own deployment. Tell us the group size and timing you need, and whether it should run inside your own perimeter.

How much does AI training for healthcare cost on ibl.ai?

There is no per-seat pricing — you pay for usage or self-host and pay only for the infrastructure, so a 5,000-person rollout does not cost 5,000 licences. 1.6M+ users across 400+ organizations run the platform this way, including NVIDIA, MIT, and Syracuse University.

Request access to Medical Coding with AI: ICD-10, CPT, and Denial Prevention

Tell us about your cohort and we will set it up — hosted by ibl.ai, or running against your own deployment, where you own all the code and the data.