Detect & classify
Find and label sensitive information in the transcript.
Healthcare Trust
SuperbCall
Privacy-first architecture for healthcare AI.
Sensitive data is redacted before any model sees it. The AI never requests clinical healthcare data — only scheduling and planning information, orchestrated by SuperbCall.
Privacy-first architecture.
SuperbCall sits in the centre and orchestrates every path — the patient conversation, redaction before the model, and scheduling integrations. The LLM never talks to clinical systems, and never receives healthcare record data.
Patient
Speaks with SuperbCall on the phone
Large Language Model
Sees redacted context only — no clinical healthcare data; replies back to SuperbCall
SuperbCall
Orchestrates the call, redaction, model, and scheduling — clear planning data stays here
Privacy Vault
Detect, mask, tokenize, minimize — redacted context only leaves here
Inside the Privacy Vault
- Detect & classify
- Filter & mask PII
- Contextual rules
- Tokenize identifiers
- Minimize data
Agenda · planning · APIs
Scheduling data only — availability and planning, never clinical records
The AI never requests clinical healthcare data. Integrations only return scheduling and planning information — always via SuperbCall. We do not store healthcare data; only what is necessary for scheduling.
Patient
Speaks with SuperbCall on the phone
SuperbCall
Orchestrates the call, redaction, model, and scheduling — clear planning data stays here
Privacy Vault
Detect, mask, tokenize, minimize — redacted context only leaves here
- Detect & classify
- Filter & mask PII
- Contextual rules
- Tokenize identifiers
- Minimize data
From SuperbCall — two separate paths (LLM never talks to agenda):
Large Language Model
Sees redacted context only — no clinical healthcare data; replies back to SuperbCall
Agenda · planning · APIs
Scheduling data only — availability and planning, never clinical records
The AI never requests clinical healthcare data. Integrations only return scheduling and planning information — always via SuperbCall. We do not store healthcare data; only what is necessary for scheduling.
AI never accesses clinical systems — or clinical data.
The language model never communicates directly with healthcare systems. It also never requests clinical healthcare data. When SuperbCall needs structured information for the call, that is limited to scheduling and planning — for example appointment availability — not diagnoses, treatments, or medical records.
Integrations are for planning only, such as:
- Agenda and scheduling software
- Appointment availability
- Opening hours and booking rules
- Practice planning calendars
- Scheduling APIs
No path from the model to clinical healthcare data. Scheduling and planning only — always through the SuperbCall backend. We do not store healthcare data; we only keep what is necessary for the scheduling conversation.
Privacy Vault.
Every transcript passes through the Privacy Vault before an AI model is invoked. Redaction happens here — not after the fact.
Filter & mask PII
Strip or mask personal data before anything reaches the model.
Contextual rules
Decide what the current task actually needs — nothing more.
Tokenize identifiers
Replace sensitive values with tokens that stay inside SuperbCall.
Minimize data
Expose only the minimum context required to continue the conversation.
The AI receives redacted context only. Clear data stays in SuperbCall’s controlled environment.
Verified scheduling data.
When the AI needs structured information, it does not fetch clinical records. It asks SuperbCall for planning data only.
The backend then:
- Retrieves verified scheduling information
- Validates responses
- Formats deterministic planning data
- Returns only what is needed for the booking conversation
Examples include:
- Available appointment slots
- Opening hours
- Bookable services (planning level)
- Reschedule / cancel availability
- Practice booking policies
Verified planning answers can go straight to Text-to-Speech — reliable for patients, with no clinical healthcare data in the loop.
Security principles.
- Privacy by Design
- Data minimization — store only what is necessary
- No storage of clinical healthcare data
- No direct AI access to healthcare systems
- No clinical healthcare data to the AI — scheduling only
- Tokenized identifiers
- Verified backend integrations
- End-to-end encrypted communication
- Complete audit logging
Designed for healthcare.
Healthcare AI needs more than conversational fluency. It needs an architecture built around privacy, security, and trust.
SuperbCall separates AI from clinical systems with the Privacy Vault and secure backend orchestration.
Conversations stay natural. Clinical healthcare data is never requested for the model or stored. Only scheduling and planning data moves through verified integrations — and we only keep what is necessary.
Building a scheduling integration?
We share architecture detail with implementers who need AVG/GDPR clarity — including what we never pull from practice systems.