Loading…
Loading…
Prior authorization decisions affect patient care and carry significant liability. Archiet generates a governed prior authorization agent: the BPMN workflow defines the clinical pathway, the DMN policy table encodes your medical necessity criteria, and the AI only extracts clinical data — it never decides the outcome.
Prior authorization (PA) decisions have three properties that make standard AI unsuitable. First, they affect patient outcomes — a wrongly denied PA delays care. Second, they must be explainable: under the No Surprises Act and CMS rules, payers must provide clinical criteria used in PA decisions, and the AMA's PA reform proposals specifically target AI-based denials that cannot be explained. Third, the clinical criteria are explicit and standard — InterQual, MCG Health, and plan-specific clinical guidelines define the conditions under which procedures are approved or require peer review. These are not learned patterns; they are written rules. Written rules belong in a DMN policy table, not a neural network.
The generated prior authorization agent implements a clinically sound workflow. Clinical data extraction: the LLM reads the PA request, clinical notes, diagnosis codes, and procedure codes and extracts structured fields — procedure category (diagnostic/surgical/therapeutic), medical necessity score (derived from diagnostic codes), plan tier, prior treatment count, and emergency flag. Medical necessity evaluation: the DMN policy table evaluates the extracted fields against your clinical criteria. Emergency procedures auto-approve. Platinum plan with high necessity score auto-approves. Diagnostic procedures with adequate clinical justification auto-approve. Low necessity scores trigger deny with explanation. The default routes to clinical review. Process routing: the BPMN process manages the clinical review queue, peer-to-peer review scheduling, appeal pathways, and regulatory timeline compliance (CMS 72-hour urgent / 3-day standard PA turnaround requirements).
The generated agent writes a complete audit record for every PA decision: the extracted clinical fields, the DMN rule that matched, the authorization decision, the clinical reviewer's identity (if applicable), and the timestamp. For denials, the audit record includes the specific clinical criterion that was not met — not a generic 'not medically necessary' statement. This audit trail satisfies the CMS PA reporting requirements, state-level PA transparency laws, and the HIPAA security rule requirements for audit controls on ePHI.
Prior authorization processing involves protected health information (PHI) under HIPAA. The generated agent includes: HIPAA-compliant data handling with encryption at rest for all PHI fields, audit logging of every PHI access event, minimum necessary data principle (the agent only extracts the fields needed for the PA decision), and a Business Associate Agreement (BAA) framework for the AI extraction component. The AI governance pack includes the HIPAA security risk assessment for the AI extraction component and the breach notification procedures.
Yes. InterQual and MCG criteria are explicit, condition-based rules — exactly what DMN tables model. Work with your clinical team to translate the relevant clinical guidelines into DMN conditions. The agent does not ship with pre-loaded InterQual/MCG content (those are licensed clinical decision support tools), but the DMN table structure is designed to implement any set of explicit clinical criteria.
The BPMN process includes a peer-to-peer review task with configurable scheduling logic. When the DMN decision routes to peer review, the BPMN process sends a notification to the requesting physician with available review slots, records the peer review outcome, and routes the final authorization decision accordingly. The entire peer review workflow is logged in the audit trail.
Confidence thresholds are applied before the DMN evaluation. If the AI extraction confidence for a field is below the configured threshold (default 0.8), the PA request is routed to clinical intake for manual field completion before policy evaluation. This prevents the DMN from making a decision based on incorrectly extracted clinical data.
The BPMN process includes an appeal pathway: when a denial is issued, a configurable window (typically 30 days) opens for the requesting physician to submit additional clinical documentation. New documents are re-extracted by the AI layer and re-evaluated against the DMN policy. If the additional information meets the clinical criteria, the denial is overturned. The appeal outcome is logged in the original PA audit record.
Most AI invoice automation is a black box — the model decides, you can't explain why. Archiet generates a governed invoice approval agent: the approval flow is BPMN (auditable), the decision logic is a DMN policy table your finance team can edit without code changes, and the AI only reads documents — it never decides the outcome.
Enterprise procurement teams need approval automation that can be explained to auditors, edited by business analysts, and trusted in regulated industries. Archiet generates a governed procurement agent: BPMN workflow, DMN policy table, bounded LLM for document reading, and a complete audit trail from day one.
Insurance claims triage requires speed, accuracy, and full auditability. Archiet generates a governed claims triage agent: FNOL intake, AI-powered data extraction from claim documents, deterministic DMN severity routing that actuaries control, and a complete audit trail for regulatory examination.
free plan. No credit card required. Generate your first compliant architecture blueprint in under 10 minutes.