Content-addressed artifacts
Every rule version has a hash, a provenance trail, and an immutable snapshot. A rule that's published in production today is the same artifact you'd see if you re-built the system from scratch tomorrow — bit-for-bit.
The strategic thesis
Three bets that change the economics.
Legacy behavioural-health RCM platforms accumulated complexity for two decades — per-customer database stored procedures, per-payer T-SQL scrubbers, brittle ETL between vendors. The new platform inverts every one of those constraints. Three architectural commitments make it work.
The architecture bet · 01
A canonical model, no stored procedures, AI-first development.
Every business rule — claim scrub, modifier injection, group-size adjustment, COB cascade, denial categorisation, EVV validation — lives in TypeScript services with content-addressed YAML rule artifacts. Eight rule-set kinds span the whole pipeline.
Scope precedence
Rules match against a 7+1 dimensional scope — org,
site, facility, billing-entity, payer, program, service-line, plus
state. Specificity scoring + precedence_rank tie-break
make the winning rule explainable, not buried in
a stored procedure.
Effective dating everywhere
Rules carry effective_from and effective_to
ranges. A fee-schedule change scheduled for Q3 doesn't require a
code deploy; an end-of-year payer policy update is a YAML change
dated for January 1.
No stored procedures. No code deploys. No back-office tickets.
The economic bet · 02
Operating leverage that doesn't grow with customer count.
The legacy model is one DBA-supported deployment per customer, one custom integration project per partner, one bespoke report per executive ask. The new platform changes the unit economics — and the headcount they imply.
$1.5–1.8k
Production Azure spend per month
Legacy on-prem equivalent: $5K–$20K/month per customer
2–3
AI development engineers
The full long-term support headcount, not a per-customer ratio
234
Production commits
In 50 business days · ~4.7 per day · 2026-04-01 → 2026-05-20
Three numbers point at the same thing — the cost structure is structurally different.
Production Azure spend is bounded by the platform's resource footprint, not by the customer count. Adding a customer adds a tenant database (a few dollars/month of Postgres storage) and a Front Door host (cents/month), not a server, not a license, not a DBA on the support rotation.
The development model is the same. Adding a state's modifier policy isn't a sprint — it's a configuration commit. The 234-commit cadence isn't a sprint either. It's the steady-state.
The compliance bet · 03
HITRUST-ready, native MFA, federation as a first-class integration model.
Security and compliance posture is built into the platform, not bolted on. The controls a HITRUST auditor would expect — they're already there, with audit evidence.
Native MFA — TOTP + WebAuthn
Phishing-resistant authentication enforceable per-tenant. FIDO MDS3 sync gates trusted authenticator models. Counter-regression detection catches credential-clone attacks. Step-up MFA on sensitive operations with a 300-second freshness window.
Federation, not custom integrations
EMR / EHR partners register themselves: RFC 7519 JWT, RFC 7517 JWK, staged → active → superseded → revoked key lifecycle. Revoke propagates across every running instance in under 30 seconds via a 30-second LRU revoke cache.
Physical tenant isolation
No row-level security. One PostgreSQL database per customer, connection strings sealed in Azure Key Vault, never persisted in the master DB. Leaking a tenant token cannot grant access to platform routes; leaking a platform token cannot grant access to any tenant's data without an explicit, audited impersonation event.
BCDR with RTO ≤ 1h, RPO ≤ 5 min
Production Postgres on ZR-HA with 35-day backup retention; geo-redundant storage (RAGZRS); a DR-by-flag posture so the passive region only costs $1.1K/month when armed. The DR drill is dated and the runbook is current.
It is structurally cheaper, structurally faster, and structurally more flexible than any legacy platform — because the architectural choices remove the categories of work that drive legacy cost.
Keep reading
Verify the thesis.
The three bets above only matter if they hold up under scrutiny. Three paths through the dossier — pick the one that matches what you want to verify.
Side-by-side
vs. Legacy
Every AdminConsole, ClientConsole, Client Central, and SuperAdmin surface mapped to its modern equivalent.
Deep diveCapabilities
The rules engine, EDI engine, master data, security model, and configuration surface — page by page.
Short formFor executives
The 10-minute read: TCO, cadence, risk reduction. Enough to forward with confidence.