Xcare: AI Operations Platform for Healthcare
- Industry
- Healthcare Operations
- Engagement
- 12 weeks · production handoff
- Client
- Xcare
Challenge
Xcare's operations team was buried under manual intake, recall, and compliance work. Senior clinicians were spending hours per week on tasks that didn't require clinical judgement.
Technical constraints
A legacy practice-management system with no modern API. PHIPA compliance requirements. Strict audit-trail obligations. Zero tolerance for hallucinated clinical data.
The legacy data problem
Patient records, appointment data, and operational state lived in a closed database designed before AI existed. Any AI automation required a safe, permissioned bridge.
AI workflow architecture
We engineered a custom MCP connector that exposed the legacy database as first-class context for AI agents — schema-aware, scoped by permission, and fully audit-logged. On top of that we shipped autonomous workflows for intake triage, follow-up scheduling, and compliance reporting.
Outcome
73% of manual ops eliminated. 4.2× throughput per operator. Compliance reporting moved from a weekly clinical burden to an autonomous workflow with 100% audit coverage.
Lessons learned
The MCP layer mattered more than the agent. Once the data was cleanly accessible, the AI workflows on top were straightforward. Most failed AI projects fail at the data layer, not the model layer.