Agentic AI Platform Inside a CQC-Regulated Clinic
A Microsoft 365-native AI agent platform for ROC Clinic, built inside their own cloud, turning Semble, Xero, HubSpot and SharePoint into skills staff can just ask.
ROC Clinic is a CQC-regulated UK private clinic offering GP, travel-clinic, screening and diagnostic services to private-pay and insured patients. It runs almost entirely on Microsoft 365 inside its own cloud tenant, with clinical, finance and operational data spread across separate systems. GreenM has an existing relationship with the clinic from an earlier data platform build.
Visit website ↗Build the platform first, so each new AI agent is a small addition rather than a new project.
GreenM is building an AI agent platform inside ROC Clinic's own cloud. Rather than one chatbot, it is a durable foundation: reusable connections to the clinic's systems, exposed as skills any agent can call, with individual agents rolled out one wave at a time.
The clinic's data and know-how sat in silos, and everyday questions and revenue-critical steps depended on individual people. Staff routed HR and policy questions back to the CEO, quotations funnelled through one person, and products administered in appointments were not always captured for billing. As a CQC-regulated provider, any solution had to keep patient data inside the clinic's own environment.
- HR and policy questions routed back to the CEO
- Quotations bottlenecked through a single person
- Revenue lost where administered products were not billed
- Clinical, finance and operational data siloed
- Strict data-residency and clinical-safety obligations
- Answers depended on who you asked
- Manual, one-person quotation process
- Billing gaps between what was done and what was invoiced
- No conversational access to the clinic's own data
GreenM designed a foundation-first platform: a searchable knowledge layer over the clinic's documents and data, a library of reusable skills that connect to each core system, orchestration on Microsoft's own agent platform with custom code where it gives more control, and a single place for staff to interact and approve. Agents ship wave by wave.
- A searchable, versioned knowledge layer over clinic documents and finance data
- Reusable skills that connect to the clinic's clinical, finance and CRM systems, with SharePoint documents synced in
- Orchestration built on Azure AI Foundry, with custom code where it gives more control
- A staff assistant in Microsoft Teams plus a Mission Control web dashboard, with a unified human-approval queue on the roadmap
- Agents delivered in waves, an HR assistant and a quotation agent first, clinical admin agents next
Because this is a regulated clinic, every consequential action, a drafted quote, a patient letter, a clinical note, needs a human to approve it before it goes anywhere. Agents are positioned as administrative assistance under human control, not autonomous decision-makers, and a unified approval queue across all agents is on the roadmap. Oversight, audit logging and privacy controls are designed into the platform from the start rather than added later.
- Reusable, secured skills shared across every agent
- Microsoft 365-native, inside the clinic's own tenant
- Human approval on every consequential action (unified queue on the roadmap)
- Versioned knowledge layer for regulated record-keeping
- Wave-by-wave rollout
The platform is in build; the foundation is in place and the first agents are already in the client's hands for testing.
A working foundation
The knowledge layer and reusable system connections are built and demonstrated inside the clinic's own cloud.
First agents in testing
An HR assistant and a quotation agent are in client testing in Microsoft Teams, with automatic draft quotes in Xero in build.
Built to expand
The skills layer is reused by each new agent, so later waves add capability without rebuilding.
Designed for compliance
Human approval, audit logging and data residency are built into the platform, not bolted on.
DELIVERY
Delivered as a platform build in defined waves, starting from discovery and a foundation layer.
Discovery & strategy
- Platform options assessed; an Azure AI Foundry plus custom-code approach chosen
- Strategy and architecture agreed with the client
Foundation & first agents
- Secure environment and knowledge layer stood up inside the clinic's tenant
- Reusable system connections built and demonstrated
- HR assistant handed to the client for testing in Microsoft Teams
- Quotation agent in client testing in Teams, with automatic draft quotes in Xero in build
Next wave: clinical admin (in progress)
- Consultation scribing: generating clinical notes automatically and writing them back into Semble
- Patient letter and report generation: drafts written back into Semble for physician review before sending
- A pre-consultation agent to prepare and streamline the pre-appointment workflow
- Microsoft Azure
- Azure AI Foundry
- Azure AI Search
- Microsoft 365
- SharePoint
- Microsoft Graph
- Model Context Protocol (MCP)
- Node.js
- Python
- Azure Cosmos DB
- Azure Blob Storage
Designed for a CQC-regulated clinic handling patient data.
Data residency
Everything runs inside the clinic's own Microsoft cloud tenant (UK South), single-tenant with private endpoints and customer-owned data stores; patient data stays in their environment.
Auditable and overseen
Consequential actions require human approval, and activity is logged for regulated record-keeping.
