AI Medical Record Review and Prior Authorization Process
A prior authorization or utilization review with AI preparing the case: clinical facts pulled from the record, matched to the medical policy criteria, approved by a nurse when they're met, and sent to a physician for any possible denial.
4 teams11 steps3 decisionsFree download
Free to download. Edit it in Flow Studio to change the teams and steps, ask AI for changes, or export to PowerPoint, Word or draw.io.
Step by step
The same flow as a numbered list, ready for an SOP, a BRD or a ticket.
| # | Step | Who | Next |
|---|---|---|---|
| 1 | Authorization request receivedStartPortal, fax or FHIR prior auth API with clinical recordsProvider · step 2 | Provider | step 2 |
| 2 | Records complete?DecisionRequired documents for the service presentReview AI · Yes: step 4, No: step 3 | Review AI | Yes: step 4, No: step 3 |
| 3 | Send the missing recordsProvider · Resubmitted: step 2 | Provider | Resubmitted: step 2 |
| 4 | Extract clinical facts with citationsDiagnoses, prior treatments, labs, imaging; each linked to page and lineReview AI · step 5 | Review AI | step 5 |
| 5 | Match to policy criteriaDocumentEach criterion: met / not met / not found, with evidenceReview AI · step 6 | Review AI | step 6 |
| 6 | Review the prepared caseNurse reviewer · step 7 | Nurse reviewer | step 7 |
| 7 | All criteria met?DecisionNurse reviewer · Yes: step 10, No or unclear: step 8 | Nurse reviewer | Yes: step 10, No or unclear: step 8 |
| 8 | Physician reviewPeer-to-peer call offered to the requesting providerPhysician reviewer · step 9 | Physician reviewer | step 9 |
| 9 | Medically necessary?DecisionPhysician reviewer · Yes: step 10, No: step 11 | Physician reviewer | Yes: step 10, No: step 11 |
| 10 | ApprovedEndProvider · Done | Provider | Done |
| 11 | Denied: can appealEndSigned by the physician; states the criteria and how to appealProvider · Done | Provider | Done |
When to use it
Use it to design AI-assisted clinical review at a health plan, hospital or review organisation, cutting time to decision while keeping clinicians responsible for every decision.
Who's involved
- Provider4
- Review AI3
- Nurse reviewer2
- Physician reviewer2
Yours works differently?
Describe your own process in a few sentences and Flow Studio draws it, with your teams and your rules.
Describe your processarrow_forwardFor the technical team
The build decisions behind this flow. Each step above also carries its technical detail.
- 01Intake: fax/portal/FHIR (Da Vinci PAS, CRD/DTR) into a case; documents OCR'd and classified.
- 02Extraction: LLM extraction of diagnoses, procedures, labs, prior treatments with page-level citations; ICD-10/CPT normalisation.
- 03Criteria: policy criteria (e.g. InterQual, MCG or the plan's own) encoded as checklists; each item met / not met / not found with evidence.
- 04Privacy: HIPAA safeguards, BAA with every AI vendor, PHI kept in your tenancy, access logged.
- 05Quality: monthly sample audits, disagreement rate between AI recommendation and clinician decision, subgroup checks.
How to get it right
- 1The AI prepares and cites; clinicians decide. Only a physician reviewer should issue an adverse decision.
- 2Show every extracted fact with a link to the exact page of the record, so reviewers can check in seconds.
- 3Watch turnaround limits: urgent requests often have to be decided within 72 hours.
- 4Audit a sample of auto-prepared cases each month for missed facts, by condition and by population.
Questions
How is AI used in prior authorization?
AI reads the submitted records, extracts the clinical facts that the medical policy asks for, matches them to the criteria and prepares a summary with citations, so nurse reviewers can approve clear cases quickly and physicians focus on the rest.
Can AI deny a prior authorization?
It shouldn't, and many rules and regulators expect a qualified clinician to make any adverse determination. Use AI to prepare and recommend, with physician review for every potential denial.
More process flow templates
- AI feature integrationTaking an AI feature from idea to production: pick the use case, test models against real examples, add guardrails, pass a security review, and roll out in stages with monitoring.
- RAG pipelineHow retrieval-augmented generation answers questions from your own documents: documents are split, embedded and indexed, then each question pulls the best passages and the model answers with citations.
- AI agent with tools and MCPHow an agentic AI system works through a task: it plans, calls tools through MCP servers with scoped access, asks a person before risky actions, checks results and loops until the goal is met.
- MCP server integrationConnecting an internal system to AI assistants through a Model Context Protocol server: design the tools, secure them, test them, review them and roll them out with monitoring.