Technological Ceiling
Set by what LLMs can and cannot do, assuming they have access to the right data, and work in a way that is safe, observable and repeatable.
Steward connects automated workflows across the health plan into one intelligence layer: alerting the right people at the right time so teams make proactive care decisions with a unified member view.
Set by what LLMs can and cannot do, assuming they have access to the right data, and work in a way that is safe, observable and repeatable.
Health plans are still designed around the constraints of human capacity and networks, often making retrospective decisions on out-of-date information, whilst optimizing siloed functions.
The current operating model of health plans cannot make use of what the models can already do. This limits AI to something that solely improves a function’s KPI and leaves member care unchanged.
Continuously interpret signal to anticipate member needs, then orchestrate interventions across members, providers, and internal teams to deliver 1:1 support at scale.
Give providers answers quickly, while making sure that they are not asked for the same information again and again.
You spend less per member: work isn't repeated across teams, and more proactive care reshapes the structure of your medical cost base itself.
Once you can steward members across functions, you can run that work for other plans. Delegated UM, care management, or similar are revenue lines a siloed organization cannot credibly sell.
Florence is our AI system designed specifically for performing highly regulated and complex work in health plan environments.
Proven across the breadth of plan workflows
Built on frontier models that can reason and work across modalities — charts, faxes, structured feeds, and policy text — so Florence handles clinical and administrative work across utilization management, case management, HEDIS, risk adjustment, and payment integrity.
Captures organizational knowledge
Florence encodes how your plan actually decides: tacit knowledge, accepted evidence, and organizational interpretations of policy and thresholds. Those standards stay version-controlled and reusable, so every workflow reflects the way your organization works.
Evaluation and feedback built in
Synthetic data and large-scale evaluation catch issues before deploy. In production, feedback from LLM checks and human review flows back into the system so Florence keeps improving with every case.
Complete traceability of reasoning
Each outcome includes the reasoning and supporting evidence behind it — relevant record findings and the specific policy criteria applied — so any result can be reviewed and substantiated on demand.
Immutable decision ledger
Every action, from record access through final outcome, is written to an immutable ledger of events. That record keeps all activity auditable for appeals, regulatory review, and internal quality assurance.
Compliant by construction
Built to HIPAA, SOC 2, NCQA/URAC, and applicable federal and state AI and interoperability requirements. Work completes automatically only where criteria are clearly met; otherwise Florence escalates to a qualified human with a sourced evidence packet.
Ingests data in its existing form
Florence uses FHIR internally and normalises inbound data on ingestion, so plans keep current formats and channels. Supported inputs include X12 EDI, HL7 v2, C-CDA, NCPDP, SFTP extracts, proprietary feeds, and unstructured PDFs, faxes, and scans.
Standard and bespoke integration paths
Documented SDKs and APIs expose Anterior's production capabilities directly for common plan systems, with clear contracts for production traffic. Bespoke integration remains available wherever a plan's architecture requires a custom path.
Flexible deployment
Interface components can run standalone or embedded in existing plan systems. Florence operates as autonomous background agents or under human-in-the-loop review, fitting the plan's operating model rather than reshaping it.