AI Agents · June 6, 2026
AI agents for medical practices: scheduling and WhatsApp
Connect WhatsApp and scheduling for a Bogotá medical practice: administrative scope, patient data, integration tests, costs and missed appointments.
An AI agent for medical practices can help answer administrative questions, check availability and coordinate appointments through WhatsApp. Its value depends on a reliable scheduling connection, practice-approved rules and a team that handles exceptions. For a practice in Bogotá, the first decision is which part of that journey needs better coordination.
When someone asks about a consultation, receives information and then waits for a manual confirmation, the gap lies between the conversation and the booking. This guide explains how to design that process, what to check before connecting systems, how to measure it and which responsibilities remain with the practice.
What an AI agent can do in a medical practice
The agent combines a language model with authorized sources and tools. The model interprets the request; a tool reads or updates the schedule. A convincing answer does not prove that an appointment has been booked: the system needs confirmation from the scheduling service and must return the date, time, professional and booking reference.
Initial tasks can cover location, opening hours, services, approved fees, payment methods, first visits or follow-ups, confirmations and rescheduling. Answers must use current information. When a fee or preparation instruction changes, someone at the practice must approve the update before the agent uses it.
Separate administration from clinical decisions
Identifying a first-visit request or a schedule change is administrative classification. Assessing symptoms, assigning medical urgency or recommending treatment is a different function. In this design the assistant does not diagnose, prescribe or perform clinical triage. Pre-appointment instructions are communicated only when approved by the practice for that service.
For clinical questions, ambiguous messages or requests for urgent care, the workflow must stop improvising and show the care channel defined by the provider. Emergency care cannot depend on a model correctly detecting a phrase or someone reading a chat promptly.
How it works with reception
The team retains complaints, exceptions, special authorizations and conversations requiring professional judgment. Define who receives each handoff, their response hours and the arrangement outside those hours. The agent must explain that a request is pending rather than imply a person has already handled it. Automation pauses when reception takes over to prevent conflicting replies.
Connecting WhatsApp to the practice schedule
The workflow starts with one authoritative source of availability. This may be a custom schedule or an existing platform, provided it supports the required operations. A WhatsApp button on a website makes contact easier; automatic booking requires an additional integration. Instagram and forms also need their own connections and permissions.
- First contact. Introduce the assistant and answer public-information questions. Ask only for data needed for the requested task.
- Availability. Check service, professional, duration, location and actual availability. Do not infer a free slot from an earlier conversation.
- Booking. Revalidate the slot when confirming. Prevent duplicate bookings when two messages arrive or an operation is retried.
- Reminder or change. Schedule contact under agreed rules. Check message status and retain a fallback when delivery fails.
- Closure. Record attendance, cancellation, rescheduling or non-attendance. A patient's confirmation message is not a completed consultation.
If the current platform lacks a suitable API, options include preparing a request for reception, linking its official booking flow or evaluating a custom schedule. The proposal must specify the option and who confirms pending requests. For channel setup, read the WhatsApp API approval guide.
Useful tests before opening the channel
Test two people trying to book the same slot, professional changes, cancellations, a temporarily unavailable schedule and a failed reminder. Also check that a request for human support stops the workflow and reaches the responsible person. Document expected outcomes before the pilot.
Patient data: define the process before automating it
Health information is sensitive data under Articles 5 and 6 of Colombia's Law 1581 of 2012. Processing is subject to restrictions and legal exceptions. The practice must establish purposes, applicable basis, access, retention and providers; a notice or checkbox alone does not settle every obligation.
As a design decision, start with the minimum information needed to coordinate the appointment. Avoid requesting medical records, test results or symptom descriptions in the administrative flow. Consider what a service name might reveal and who can see notifications. Credentials, permissions and access logs belong in the implementation, alongside a process for handling personal-data requests.
The WhatsApp Business policy makes use of health information conditional on applicable requirements. It also requires permission for subsequent contact, honoring opt-outs and approved templates outside the 24-hour service window. Review these conditions before enabling reminders or campaigns.
Measuring missed appointments without promising a recovery rate
First define non-attendance and its denominator. One measure is appointments missed because the patient was absent divided by appointments due in the period. Report cancellations and rescheduling separately, using consistent rules to avoid duplication. Segment by service, professional, location and booking lead time when volume permits.
Compare similar periods and record changes to reminders, cancellation rules, availability and inquiry sources. The guide to no-show strategies explores possible actions; this guide focuses on the integration needed to record and execute them. A subsequent improvement does not establish AI as its sole cause.
| Indicator | Required record | Decision it supports |
|---|---|---|
| Pending inquiries | Arrival, response time and owner | Improve coverage or fix handoffs |
| Recorded bookings | Reference confirmed by the schedule | Find friction between inquiry and booking |
| Reminder delivery | Channel status and any failure | Retry under agreed rules or use a fallback |
| Attendance and cancellation | Final state and time of change | Evaluate occupancy and released slots |
What the budget should separate
Separate implementation, monthly operation and third-party usage. Setup should describe the service catalog, schedule, channel enablement, rules, testing and training. Operation should define support, incident review and included adjustments. Specify how Meta messaging, AI model usage, infrastructure and additional tools are charged.
The Consultorio IA catalog lists price in your MD account for setup and price in your MD account for monthly operation to signed-in users. Review the current Consultorio IA scope and request a proposal confirming your practice's integrations. An implementation fee does not guarantee patient volume or a particular return.
Implementation and choosing a provider
Start with one service and a bounded journey. Document the catalog, schedule, owners and handoff rules; enable accounts and access; test with fictional data; run a supervised pilot; then expand when records show the workflow is working. Meta approval and API availability are dependencies that belong in the timeline.
Before contracting, ask to see availability, booking and a controlled failure. Ask how double booking is prevented, what happens when someone requests a person, how actions remain traceable and who updates information. Check the exit process too: data access, export, credentials and termination procedures.
The healthcare demonstration lets you explore a simulated scenario; it does not verify a connection to your schedule. To evaluate your operation, let's discuss your services, inquiry volume and current system. Those inputs support a reviewable implementation plan and budget.