Voice-Bot Appointment Booking for Clinics: DPDP Consent at Every Step

Why a patient call is personal data, not metadata

When a patient calls a clinic and says 'I have been having severe chest pain since this morning,' the clinic is now in possession of sensitive personal health information. The Digital Personal Data Protection Act 2023 does not distinguish between data captured by a form and data captured by a voice recording — both are personal data, and the more sensitive the data, the higher the regulatory bar. Clinics that record patient calls without explicit consent are creating an audit exposure that will not be visible until the first complaint.

Consent at the greeting — not at the goodbye

The pattern we have locked in across nine clinic deployments: the voice bot's opening line includes a one-sentence consent prompt before any health information is solicited. 'Hello, you have reached Sai Clinic. This call may be recorded for appointment-booking purposes; we will not use this recording for marketing. Press 1 to continue, or stay on the line if you'd like a human to call you back.' This single-sentence pattern preserves DPDP consent without breaking the user experience.

• Recording-consent prompt before any health information is solicited
• Purpose limitation: only appointment-booking — never marketing or research without separate consent
• Explicit human-fallback option at the greeting (not buried later)
• Audit log captures consent timestamp, language used, and the bot's exact wording
• Re-consent prompt for callers re-confirming an appointment more than 12 months later

Symptom capture: when to stop the bot

A voice bot taking an appointment should ask 'which speciality?' or 'which doctor?' — not 'what symptoms are you having?' The moment a patient begins describing symptoms in detail, the bot is straying into triage territory that it is not licenced or capable to handle. The script we have settled on: the bot accepts a one-sentence symptom mention (because some patients lead with it), confirms the speciality match, and books the appointment without probing deeper. Symptom probing is the doctor's job, not the bot's.

Triage vs booking: a line the bot should not cross

A voice bot that crosses from booking into triage is creating both a clinical risk (advising the wrong specialty) and a regulatory risk (medical device classification could apply). The line we recommend: the bot books appointments and handles cancellations; anything that involves a clinical recommendation goes to a human. If a patient mentions chest pain, severe bleeding, breathing difficulty, or anything else with an emergency signature, the bot must immediately route to a human nurse triage line — and have the script for this be unambiguous. The bot's restraint here protects both the patient and the clinic.

PHI redaction before any AI model sees the transcript

If the voice bot transcript is processed by a language model for summarization or categorization, the transcript must be redacted of Protected Health Information first — patient name, phone, address, Aadhaar, ABHA ID, and any free-text mention of family relationships. This is not just good practice; it is a foundational pattern for any DPDP-aware build. The redaction should happen inside the clinic's controller perimeter — never relying on a third-party model to enforce its own redaction.

The audit trail that survives a DPDP inquiry

When a DPDP complaint is filed against a clinic, the clinic's defense is its audit trail: the consent prompt the patient heard, the patient's response, the data points captured, the human handoffs, and the retention period applied. A voice bot deployment that does not produce a queryable audit trail for every single call is not deployment-ready. The audit log should be append-only, retained for the regulatory minimum (three years for ordinary records, six for paediatric and oncology), and exportable to inspectors in a single click.

A healthcare voice bot that does not pause for explicit consent is a healthcare voice bot that has not understood the DPDP Act.

Internal DPDP readiness review, 2024