Network Management Agentic Workforce.
A PE-backed multi-state behavioral health provider group went from no contracting infrastructure to a Claude-run contracting engine in three weeks: 1,000+ documents read, 82 contracts normalized to one schema, and a $7M opportunity identified across the full contract book.
Company Profile
A PE-backed provider of behavioral health care to residents of nursing and assisted living facilities in four states: New Jersey, New York, Pennsylvania and Maryland. After a 2024 private equity investment it added two companies and ran as three legacy entities, each with its own payer contracts across Medicaid, Medicare Advantage and commercial lines. No centralized contracting function, no shared repository, and no analytical layer on top of rates, codes or renewal calendars.
The Challenge
Contracts, amendments and fee schedules sat in shared folders in many formats, some held by an outside vendor, and nobody kept a structured record. Leadership could not see which contracts were in force, when notice windows fell, what each payer paid across entities and states, or which documents were missing.
That limited the financial upside on the current book, left the operational team without the resources to work it, and capped growth: the company planned to expand as Medicare Advantage grew in its markets, and it needed a clear view of its payer book first.
The PE sponsor’s investment thesis depended on commercial scale across the three entities. The commercial engine could not scale on a foundation of PDFs and spreadsheets.
The 3-Week Engagement
3Pillars stood up a network management agentic workforce on Claude: a commercial contracting engine built from a standing start. In health plan terms, payer contracting is network management, so the design was modeled on how a regional health plan runs its contracting shop and sized for a multi-state provider group.
Week 1: Diagnostic
Sessions with revenue cycle, finance and the leaders of each legacy entity. A sweep of shared drives and email archives produced a corpus of more than a thousand files. Before the full build, the agent design ran end to end against that corpus: sorting contracts from non-contracts, identifying payers, normalizing carrier families across M&A, flagging gaps and producing a first-pass outreach list.
Weeks 2 to 3: Build
Extraction agents in Claude Cowork read every contract, amendment and fee schedule and normalized each one into a 42-field record with dates, signers and CPT-level rates, held in the client’s own AWS environment. 3Pillars reviewed the extractions. Analysis agents in Claude Code queried that database through MCP and wrote seven briefs: renewal risk, payer mix, rates against the book and 2026 Medicare, rate-increase terms, value-based terms and entity differences.
The engine reads vertically inside a single payer relationship and horizontally across payers, entities and states. Three legacy entities meant three tax IDs, three ways of describing the same service and three sets of rates; the build reconciled them so the agents could compare like with like. 3Pillars then sent payer document requests and presented a readout with 13 ranked initiatives.
Results
On the 8 priority payers alone (about $25M of trailing revenue) the conservative annualized upside is $2.0M to $4.5M+, or 8% to 18%. The renewal matrix listed 104 outreach items across 46 contracts; 12 renewal-date gaps were closed and 15+ missing contracts and fee schedules sourced. Phase I is complete.
Functional Areas Touched
What started as a contracting build rippled across the organization. The map below shows a typical risk-bearing provider operating structure. Highlighted nodes are where this engagement created downstream impact.