Solutions / Prior authorization
Healthcare and benefits
Prior authorization without the fax machine
Requests assembled from the chart, matched to the payer’s criteria with citations, submitted, tracked and appealed, with a clinician approving every submission.
Representative blueprint. The targets below are design targets we scope and measure against in a Loop Audit. They are not results from a named client.
The administrative burden hiding inside every treatment plan
- Each payer has its own criteria, forms and portals, and they change without notice.
- Staff re-read the same chart to find the one sentence that satisfies the criteria.
- Requests bounce for missing information that was knowable on day one.
- Denials are appealed late, or not at all, because nobody has time to write the letter.
- Patients wait, and care is delayed.
How the loop runs.
7 stages, from the moment work arrives to the moment it is proven done. Each one is observable, testable and can be paused.
Detect
Spots orders that need authorization from the scheduling and EHR feed, and identifies the correct payer and plan.
Gather
Collects the clinical notes, labs, imaging reports and prior treatments relevant to the request.
Match criteria
Compares the chart to the payer’s current criteria and shows, for each criterion, the exact passage that supports it or the gap.
Close gaps
Asks the clinic for missing documentation before submission instead of after rejection.
Clinician review
Presents the assembled request with citations. A clinician approves, edits or rejects.
Submit and track
Submits through the payer’s portal or API, captures confirmations, polls status and alerts on changes.
Appeal
On denial, drafts an appeal that cites the criteria and the chart, prepares peer-to-peer talking points, and tracks the appeal clock.
The hard parts, and how we handle them.
Criteria are living documents
Payer criteria are stored as versioned libraries with effective dates. A change creates a diff and a review task, so a stale rule never silently drives a submission.
Every claim in a request must be traceable to the chart
Each statement links to its source passage. If it cannot be cited, it is not written. This is how we keep a language model honest in a clinical setting.
Portals do not have APIs
Where a payer only offers a portal, we automate the browser with screenshot evidence at each step, resilient selectors, and a human fallback when the portal changes.
Protected health information everywhere
Minimum necessary access, field-level redaction in model calls, region-pinned processing, and no retention of PHI by model providers.
What a person always decides
- A clinician approves every submission and every appeal
- No clinical judgement is made by the system, only assembly and citation
- Model providers are contractually barred from retaining or training on PHI
- Access is role-based and logged at the record level
- Automatic pause if a payer portal behaves unexpectedly
Connects to
- Epic
- Cerner
- Accuro
- OSCAR
- FHIR APIs
- Payer and insurer portals
- Secure fax and e-fax gateways
Built with
- Chart extraction with passage-level citations
- Versioned criteria library
- Browser automation with evidence capture
- Case tracker with deadline engine
- FHIR and HL7 integration layer
A realistic timeline.
Shadow mode first. Nothing acts on its own until its accuracy has been measured against your team’s own work.
- Weeks 1 to 3Map the top authorization types by volume and denial rate. Collect payer criteria. Privacy and security review.
- Weeks 4 to 8Gather, criteria matching and citation review with staff, on past cases first.
- Weeks 9 to 12Live submission for the first payers behind clinician approval. Status tracking on.
- Week 13 onwardAppeals drafting, more payers, denial analytics.
What we design for.
Modeled targets, confirmed or corrected with your own data in a Loop Audit.
- Designed for PHIPA and HIPAA obligations
- Data residency and BAA-style terms on request
- Full access and decision audit trail