AI PBM Contract Negotiation Prep Agent

How Cassidy automates this using AI

Step 1: Trigger on contract cycle or data upload

The Workflow activates when a new PBM proposal arrives, claims data is uploaded, or a scheduled market check begins pulling in adjudicated pharmacy claims, current contract exhibits, formulary files, and rebate schedules.

Step 2: Normalize and map pricing benchmarks

Cassidy harmonizes drug pricing data across AWP, WAC, and NADAC benchmarks, then maps claims to channel and day-supply categories (retail 30/90, mail, specialty) to establish baseline PMPM, generic dispense rate, and specialty mix.

Step 3: Parse contract definitions and flag risks

The agent extracts and compares critical definitions brand vs generic classification, specialty thresholds, lesser-of logic, MAC methodology, and audit rights identifying cross-subsidization risks and gaps that could affect guarantee settlement.

Step 4: Benchmark guarantees and model net cost

Cassidy evaluates AWP discount guarantees, dispensing fees, and per-claim rebate structures against market ranges, then builds a net cost model (ingredient cost minus discounts plus fees minus rebates) with sensitivity analysis for mix shifts and new drug launches.

Step 5: Reconcile rebates and detect spread pricing

Automated analytics compare quarterly rebate remittances to expected accruals by NDC and class, while spread pricing diagnostics infer gaps between plan-paid costs and pharmacy reimbursement to surface hidden PBM margin.

Step 6: Generate redlines, RFP artifacts, and scorecards

Cassidy drafts contract redlines enforcing pass-through pricing, 100% manufacturer-derived revenue sharing, and dollar-for-dollar guarantee settlement, plus standardized RFP grids, BAFO templates, and bidder comparison scorecards.

Step 7: Deliver negotiation playbook for review

The completed package net cost models, redlined terms, rebate reconciliation, and prioritized negotiation levers with quantified savings routes to benefits leaders via Slack, Teams, or email with Human-in-the-Loop approval before finalizing.