The Platform
The Intelligence FabricA domain-aware agentic AI platform for hospital operations
A network of purpose-built, semi-autonomous AI agents that read your clinical and operational data, identify what needs to happen, and support your teams to act — with human oversight built into every consequential step.
- Kubernetes-native (on-prem / cloud)
- Multi-cloud portability
- Air-gap & hybrid support
- Lifecycle management
- Multi-agent orchestration
- HITL governance
- Self-learning (RLHF/GEPA)
- HL7, FHIR, REST, CDC
- EMR / EHR, HIS connectors
- CRM, SCM, lab systems
- Domain-specific data cubes
- Cross-domain federation
- Governed metric definitions
Platform Architecture
Built in layers. Deployed in weeks.
The Intelligence Fabric is structured as four distinct layers — from data ingestion at the bottom through to domain agent applications at the top. Each layer is independently deployable and connects cleanly to the next. Click any layer to see the full detail.
Six domain-specific agent types operate across the three platform applications: Patient Engagement, Care Continuity, RCM Optimisation, Supply Intelligence, GRC Compliance, and Analytics & Insights. Each agent is configured against validated operational bottlenecks identified during the audit phase.
The core of the platform: the agent runtime, orchestration layer, and human-in-the-loop governance framework. Key capabilities:
- Agent lifecycle management — deploy, monitor, retrain, and retire agents systematically
- Multi-agent orchestration — agents that coordinate across domains (e.g., a care gap agent triggering a follow-up agent)
- Human-in-the-loop (HITL) governance — every agent has defined confidence thresholds and impact boundaries above which a human must confirm
- Self-learning (RLHF/GEPA) — agents improve from operational feedback, exception analysis, and outcome data
- LLM observability — full visibility into model behaviour, prompt performance, and output quality
- Cost-aware model execution — task-specific model routing to balance accuracy and cost
- Conversational mode — natural language query across operational data cubes, beyond fixed dashboards
Raw system data is cleaned, normalised, and structured into domain-specific data cubes — clinical, financial, operational. Governed metric definitions. Historical and real-time views. Cross-domain federation for workflows that span departments. This is the factual foundation agents reason over.
The Intelligence Fabric connects to your existing systems via HL7, FHIR, HL7 v2, REST API, or flat file. Pre-built connectors for major EMR, HIS, CRM, and SCM platforms. File, API, and CDC ingestion. Real-time and batch pipelines. Your data stays in your environment.
Agents
Six agents across three domains
Each agent is purpose-built for a specific operational domain and configured against the validated bottlenecks identified during your Audit & Discover phase. No generic assistants — every agent has a defined role, defined confidence thresholds, and defined escalation rules.
Patient Engagement Agent
Manages outreach queues for acquisition leads and retention follow-ups. Triggers communications at the right point in the patient journey. Tracks responses and escalates non-responders to phone-based follow-up.
Care Continuity Agent
Reads discharge summaries, prescriptions, radiology reports, and lab results to identify follow-up requirements. Generates structured care gap records for the retention team's review.
RCM Optimisation Agent
Supports claim review, denial prediction, and coding assistance. Reviews assembled claim packages against payer rules before submission. Identifies patterns in deduction and rejection data.
Supply Intelligence Agent
Analyses demand signals across clinical departments. Supports procurement orchestration and inventory optimisation. Surfaces forecasts for human procurement team review and approval.
GRC Compliance Agent
Monitors policy adherence, flags risk, and supports audit trail preparation. Operates continuously — not on a monthly or quarterly cycle. Human compliance team reviews flagged cases.
Analytics & Insights Agent
Natural language query across operational data cubes for any domain. Enables conversational drill-down into clinical, financial, and operational data — beyond fixed dashboards.
Governance Model
Every agent has a human in the loop. Here's what that means in practice.
Semi-autonomous means agents do the work that doesn't require human judgment, and surface the decisions that do — with the context needed to make them well.
Agent handles autonomously
Continuous monitoring, KPI computation, pattern detection, routine alerting, scheduling reminders, feedback routing, data quality monitoring. No human needed in the loop for each individual event — only for the exceptions.
Agent handles, human validates
Ambiguous guideline interpretations, high-impact configuration changes, novel failure modes, clinical content for patient communications, write-off approvals above defined thresholds. Agent does the work and presents a bounded decision for human confirmation. Typically 15–40 minutes of human time per occurrence.
Human-led, agent supports
Payer relationship negotiation, automation roadmap strategy, process redesign, staff adoption and change management, accountability for consequential errors. Agent prepares the evidence and analysis; human leads the decision and the action.
The practical result: your team's time shifts from manual data-chasing and routine tracking toward high-judgment oversight of a system that handles the operational load.
Infrastructure
Deployed in your environment. Not ours.
On-Premise Kubernetes
Full deployment within your own data centre. Air-gap capable for maximum data sovereignty. Patient data never leaves your infrastructure.
Google Cloud
Managed cloud deployment with customer-controlled access and data residency. Deployed in your GCP project — not HealthFoundry's.
AWS / Azure
Standard cloud deployment options with your existing cloud agreements and IAM policies. Kubernetes-native — portable across providers.
Hybrid / Air-Gapped
For environments requiring split deployment across on-premise and cloud components. Designed for the most stringent data residency requirements.
The Intelligence Fabric is Kubernetes-native — designed for portability across any compliant infrastructure. Your patient data never leaves your environment without your explicit authorisation.
AI Engine
Models built for healthcare, not adapted from general purpose.
The Intelligence Fabric uses a combination of model types, selected by task and optimised for cost and accuracy. Model selection is cost-aware — the platform routes each task to the right model at the right cost, avoiding LLM overhead for deterministic tasks.
Large Language Models
For document understanding, clinical content extraction, and natural language query. Fine-tuned on healthcare terminology and workflow context.
Deep Learning Models
For pattern recognition across operational and financial data — denial prediction, care gap identification, supply demand forecasting.
Task-Specific Fine-Tuned Models
Fine-tuned on customer data for domain-specific accuracy — coding validation against ICD-10 and payer package definitions, care gap extraction, deduction analysis.
RLHF / GEPA
Reinforcement learning from human feedback, driving continuous improvement from operational use. Agents improve from staff feedback, exceptions, and outcome data — automatically, without manual retraining cycles.
Ready to see the Applied Intelligence Platform in your environment?
Engagements begin with a structured discovery session where we map your current operations, validate your KPIs, and show you where agents can be deployed first for the fastest measurable outcome.
Book a Call