OPERATIONAL CONTEXT GRAPH · v1.11 · LIVE

Infrastructure for healthcare AI.

One operational graph underneath every healthcare answer.

Generating a healthcare agent is the easy part. Grounding it in 1,958 mapped processes — and the regulations, standards, and CMS benchmarks behind each — is the part Amandil already built. Build agents you own, run them in your stack, kept current as healthcare moves.

THE GRAPH

What the graph knows. What it governs.

Two sides of one graph: the knowledge an agent reasons from, and the guardrails that govern what it's allowed to do.

01 · KNOWLEDGE

What OCG knows about healthcare operations.

Knowledge grounds reasoning. Every process linked, every metric cited.

22
DOMAINSRevenue Cycle, Patient Access, Compliance + more
194
SUBDOMAINSeach holds 8–12 processes with owners & metrics
1,958
PROCESSESatomic units of work — scored, linked, citeable
10,140
RELATIONSHIPStyped edges wiring processes across domains
6,500+
CITED METRICSindustry benchmarks sourced to public filings
260
FTE-DAY MODELscoring baseline for automation feasibility
+
02 · GUARDRAILS

What governs agent action.

Guardrails govern action. The graph defines the bounds; the agent enforces them.

2,493
REGULATIONSHIPAA · CMS · NCQA
2,201
STANDARDSX12 · FHIR · HL7 · CPT · ICD-10
847
POLICY CLAUSESpayer rules — prior auth, medical necessity, coverage
312
AUDIT CHECKPOINTSrequired evidence & attestations
58
ESCALATION PATHShand-offs when confidence or policy fails
100%
TRACE COVERAGEevery agent action grounded to a cited source
AGENTS · ARCHETYPES

Every process can host an agent. Three examples.

1,958 processes in the healthcare graph — any one of them can host an agent that cites the regulation, applies the standard, and returns an audit trail. The three below illustrate the shape the substrate produces — pattern, mechanism, output. Design partners build the first production deployments.

A.01EXAMPLE
Prior-Auth Reconstructor
Rebuilds the missing context behind every denied prior auth — payer policy, NCD/LCD, prior-auth criteria, member benefits, prior decisions — and drafts the resubmission with citations.
D.01 · D.04 · D.09Focus · prior-auth cycle
A.02EXAMPLE
Denial Pattern Analyzer
Reads denied claims against the graph, classifies denial root cause (coding, eligibility, medical necessity, timely filing), and routes to the responsible workflow with the rule and citation.
D.01 · D.02 · D.04Focus · denial recovery
A.05EXAMPLE
Medicaid Redetermination Compiler
Walks every enrollee against the HR1 six-month redetermination cadence, runs work/community-engagement and income re-verifications against source records, drafts member outreach with the rule cited, and flags coverage at risk before the disenrollment trigger.
D.04 · D.07 · D.11Focus · coverage retention
1,958 PROCESSES — EACH A POSSIBLE AGENT USE CASE NOT LISTED? REQUEST IT →
WHO IT'S FOR · HEALTHCARE

Who runs on the graph. The same three customer segments.

The same three customer segments that anchor every Amandil vertical. Payers and providers run the operations. Consultancies compress discovery into branded analysis. Health-IT vendors embed the graph into their products.

A · PAYERS & PROVIDERS

Where should we invest in AI?

Show me the processes CMS thinks we're worst at — ranked by automation feasibility and payback.

Sixteen of twenty-two domains are shared between payer and provider — the mirror_of relationship pairs both views of the same work so the seam stops being a black box. Diagnose, benchmark, simulate, automate.

  • CMS benchmarking, facility & system
  • Full payer operations + automation scoring
  • Simulation-backed business cases
  • Mirror_of pairs payer↔provider views
B · CONSULTING PARTNERS

How do we scale healthcare AI engagements?

Replace six weeks of discovery with a 30-minute branded analysis.

Start at the end state. The graph already contains every domain, process, standard, regulation, role, and benchmark. Walk into the first leadership meeting with a ranked list of opportunities and a defensible citation for each.

  • White-label per client · brand, colors, exports
  • Engagement-ready gap analysis from CMS profile
  • Branded PDF blueprints, business cases, diagnostics
  • Multi-client workspaces with scoped grants
C · HEALTH IT VENDORS

What grounds our healthcare AI features?

Inherit a healthcare operations context layer instead of building one.

Embed Amandil through a native Model Context Protocol server. Issue a bearer token, scope to allowed tools and L1 domains, and your product inherits 1,958 grounded processes, 50+ standards, 20+ regulatory frameworks — all citeable, all versioned.

  • MCP server · 13 tools · 3 read-only resources
  • OAuth 2.0 + headless bearer tokens, instantly revocable
  • Grant-scoped: tools and L1 domain prefixes per partner
  • Versioned surface area · stability for build-on partners
REQUEST ROADMAP

See your gap. No data required.

No data sharing, no integration. We benchmark your organization against CMS public data — and hand back the processes to fix first.

WHAT YOU RECEIVE · 5 BUSINESS DAYS

A gap report drawn from your public CMS footprint — plus the processes to fix first.

Annual platform licensing. Available via app, API, or AI assistant integration. Live demo against the graph included in every roadmap request.

DeliveryGap report · top 10 targets · blueprint sample
DataCMS provider PUF · system-level benchmark
AccessApp · API · MCP · Claude.ai
COMMON QUESTIONS

Questions, answered.

Is Amandil's healthcare Operational Context Graph live?

Yes. It's live at v1.11 — 22 domains, 1,958 processes, 10,140 typed edges, stitched to public CMS data for 9,000+ hospitals, with 2,493 regulations reverse-indexed. Every node is citeable and versioned.

Does Amandil ingest our private patient data?

No. The graph is a public reference architecture of how healthcare operates — its processes, standards, and 2,493 regulations — with peer benchmarks drawn from public sources such as CMS and HFMA. None of it is your internal systems. Benchmark against the public data, or drop in your own; no integration is required to start.

What standards and regulations does it cover?

CMS Medicare data, HIPAA, NCQA, and Joint Commission, plus reference standards X12, FHIR, HL7v2, CPT, HCPCS, and ICD-10 — with 2,493 regulations reverse-indexed to the processes they govern.

How do AI agents access the graph?

Through native Model Context Protocol (MCP) for Claude, Cursor, and any MCP-compatible framework, or REST for legacy stacks. The agent calls the graph like any other tool and gets structured, token-budgeted context with a citation on every answer.