Ahead of the Clawback
AuditAhead builds board-admissible, cryptographically signed evidence for every pharmacy claim, using four AI agents that cross-check each other, while a licensed pharmacist remains the only one who can sign off on a submission.
Most pharmacies don’t know their exposure until the recovery letter arrives.
A 100-store Ontario chain carries an estimated $2.85M in annual fiduciary exposure, based on the province’s 2024 Auditor General report on program-wide billing. That body of audit work points to four recurring documentation failures behind nearly every clawback.
- Missing or unsigned consent.
- Eligibility mismatches.
- Services that cannot be substantiated in the clinical notes.
- An incorrect drug, DIN, or quantity code on a claim.
Publicly reported recoveries at individual pharmacies have run from the high hundreds of thousands to several million dollars, and compliance staff already spend upward of twelve hours a week on this kind of documentation review, per the sector’s 2023 regulatory burden research. Catching these failures has so far depended on that manual review keeping pace with claim volume.
The design question: how to catch a documentation failure before a claim is submitted, in a form a pharmacist can trust enough to sign, and a Ministry auditor can trust enough to accept.
A planning model, not a delivered result.
Read what’s already there. Cross-check like an auditor. Sign only when a person signs.
Read what a pharmacist already produces
Parse consent forms and clinical documentation as they exist today, including handwritten or non-standard layouts, using OCR and AI trained per format, rather than requiring a new workflow.
Cross-check claims the way an auditor would
A claims analyst agent reads the plain-English Limited Use criteria and checks billing against the prescription, then a separate forensic reviewer agent independently challenges that output and flags edge cases, an adversarial design rather than a single pass.
Sign, don’t submit, until a person does
Every batch produces a cryptographically signed digest written to an append-only ledger, but nothing reaches the Ministry until a licensed pharmacist reviews and countersigns it.
Four agents, one countersignature.
Consent and PHI boundary
Personal health information is redacted before it crosses to the next agent.
The first agent parses consent documentation and redacts personal health information before it crosses to the next agent, so no raw identifiers reach the claims analyst.
Claims cross-check
Billing is checked against the prescription and formulary rules, in plain English.
The claims analyst agent reads Limited Use criteria in plain English and cross-checks billing against the prescription and formulary rules.
Adversarial review
A dedicated agent independently challenges the other agents’ output, rather than a single pass.
A dedicated forensic reviewer agent independently challenges the output of the preceding agents and flags edge cases, rather than the batch simply passing through a single review.
Signed evidence
Every batch gets a cryptographically signed digest in an append-only, ten-year ledger.
The final agent signs a SHA-256 and HMAC-SHA256 digest for the batch and writes it to an append-only, ten-year ledger.
Human-in-the-loop gate
Continuous monitoring alerts the audit committee before a recovery letter would arrive.
Continuous monitoring fires an alert if the discrepancy rate crosses a configured threshold across multiple batches, notifying the audit committee before a Ministry recovery letter would arrive. The system never auto-approves a submission.
Signed, not just stored. Gated, not automatic.
A pharmacist remains the legal signer
The system never auto-submits to the Ministry. A licensed pharmacist’s countersignature is required on every batch, and the accountability log is time-stamped.
Evidence is signed, not just stored
Every batch gets a SHA-256 and HMAC-SHA256 signed digest in an append-only ledger, verifiable offline by an external auditor, with tenant-owned signing keys.
PHI never crosses agent boundaries unprotected
Personal health information is redacted structurally before it reaches the claims analyst agent, and processing runs under a signed information management agreement in the Canadian region.
From configurable thresholds to cross-province expansion.
Customization: Configurable discrepancy thresholds (2%, 5%, 20% per pharmacy), adapting instantly.
Expansion: Drug-benefit audit attestation, piloted at a single large Ontario chain.
Customization: Custom field and document mapping, with new formats trained into the parsing agent.
Expansion: Medication-review and banner rollout across multiple chains on the same architecture.
Customization: Multi-system integration by FTP, API, or folder sync, on an open-standards architecture.
Expansion: Minor ailments coverage and cross-province expansion.
Customization: Per-agent rule tuning: formulary packs, LU criteria, and consent standards configured independently.
Expansion: Long-term care, federal, and adjacent industries, with per-tenant isolation that scales by volume.
Four steps from approval to demo.
Pilot approval
Approve the pilot scope and designate an executive sponsor; integration begins within two weeks of approval.
Real data access
Share anonymized batch data so the system can be tuned before go-live.
Production cost modelling
Finalize cost-per-document tiers (1K, 10K, and 100K documents a month) to design the subscription model.
Stakeholder demo
A live walkthrough with pharmacy operations and the compliance director, using real batch data.