Back to blog
Medical Tourism Operations

Medical tourism: how to test a WhatsApp chatbot before it handles leads

A convincing reply can hide an incomplete intake. Build medical tourism test scenarios in AriaBee, control live outbound messaging and inspect the evidence before handling patient enquiries.

•Published: •Updated: •9 minute read

To test WhatsApp automation before it handles medical tourism leads, define the expected outcome for realistic patient enquiries, run controlled conversations and check the replies. Verify claimed actions only where the product provides supporting evidence; otherwise, mark them as unverified. In AriaBee, use automated scenarios with a synthetic lead model, keep live outbound WhatsApp permission off during simulation, then review each score, turn count and message log. Include relatives enquiring for patients, missing identity documents, medical uploads, language changes and budget objections.

WhatsApp is where the enquiry and its documents arrive. For a medical tourism agency, the work includes establishing who needs treatment, collecting the right files and coordinating the next step. AriaBee is the AI operations employee for medical tourism agencies — commanded from WhatsApp. Staff delegate ordinary work by text or voice; managers use the web interface to configure behavior, review evidence and handle exceptions.

A friendly reply can still leave the patient intake wrong

Consider this illustrative opening: “My mother needs treatment in Istanbul. I have her reports, but I don’t have her passport yet. Can we arrange a call?” An agreeable response might sound helpful while treating the sender as the patient or offering a booking before your agency’s required identity check.

The operations manager then has to untangle the relationship, establish whose documents arrived and explain why the next step is delayed. The owner pays through repeated coordinator work, incomplete records and a family being asked to provide information again. Across the wider journey, incorrect identity and travel details can contribute to rejected visas, missed flights and mis-scheduled appointments.

A chatbot test that stops at “the answer reads well” leaves that operational work unchecked. Wati’s testing guide describes manual conversations from a personal or test WhatsApp number to verify responses, triggers and flows. That establishes conversational behavior. Medical tourism readiness also requires checking patient identity, prerequisites and recorded outcomes. Waaru’s testing guide makes the useful distinction between a convincing reply and evidence that the underlying business action succeeded.

Set up AriaBee’s scenario tests before running conversations

Our August 2, 2026 release, “Scenario Runner & Dynamic Fields,” introduced automated scenario test runners, debug inspector panels and dynamic field cards. The documented runner uses a synthetic lead AI model to simulate conversations and score performance and flow. It gives the manager a repeatable way to examine cases that an ordinary demonstration can skip.

  1. Open AI Sales Agents from the left navigation, select the relevant agent and open Test.
  2. Open the test configuration from the scenario page header. Choose the model that evaluates conversations and the default model that plays the synthetic lead.
  3. Configure the scoring rubric and severity levels around your expected outcomes: correct patient identification, respect for document requirements, factual package answers and appropriate human involvement.
  4. Check the test-runner permission for live outbound WhatsApp sending before starting any run. Keep that permission off for simulation.
  5. Create a scenario or edit an existing one. Configure the simulated client’s name, language, opening message, personality parameters, treatment goals, budget concerns and stop conditions. Attach made-up medical files where the case requires them.

Check live outbound permission explicitly

The test configuration includes a flag that permits live outbound WhatsApp messages. Enabling it allows real outbound sending; a synthetic lead does not by itself establish that no live messages can leave. The help guide does not state the default. Verify the setting before each controlled session and enable it only for a deliberate delivery test using a contact your team controls.

Write the expected outcome before running the case. Set a stop condition that leaves enough conversation to observe it. For a missing-document case, successful handling may mean explaining the requirement and keeping the next action on hold. A completed booking would be the wrong result if the configured prerequisite remains unmet.

Build the test matrix around medical tourism exceptions

Use the following cases as an illustrative starting set. Adapt them to your agency’s actual policies and enabled capabilities. These are proposed checks, not claims that AriaBee has passed your configuration.

A relative is enquiring for the patient

Start with: “I’m asking about treatment for my mother.” Enable Ask who is being treated in Capabilities. Expect AriaBee to establish the treatment subject before checking or booking call availability. Inspect the conversation for any assumption that the sender is the patient. Our September releases added caller-identity safeguards and explicit patient-status confirmation before appointment offers.

A required identity document is missing

Have the simulated lead ask to proceed without identification. Test each requirement your agency actually enables: Require patient ID before a sales call, Require patient ID before doctor review or Require patient ID before a price quotation. Expect the corresponding action to remain held. Check that the explanation matches your configured policy. The runner documents medical-file attachments; this missing-ID case does not require assuming it supports simulated identity-document uploads.

Medical files arrive during the conversation

Attach a made-up report to a scenario and enable File Intake. Define what should happen after receipt, including the configured pipeline stage and any remaining intake questions. Review the acknowledgement in the message log. Verify any claimed pipeline change only where the product provides supporting evidence; otherwise, record it as unverified. A message saying the report arrived does not establish that a doctor reviewed it or that treatment is appropriate. AriaBee does not replace medical judgment.

The conversation changes language

Set the scenario’s starting language, then include a language-change request in a controlled conversation. Review Auto-detect Language under Agent Profile. Inspect the turns around the switch: did AriaBee complete the confirmed change, preserve the patient context and continue the outstanding task? Our September 28 release added language-state tracking to complete confirmed switches. See the wider workflow in multilingual patient communication.

The patient objects to the budget

Use the scenario’s budget concerns to challenge a package price: “That is above what we can spend. Can you guarantee the result if we pay?” Expect an answer grounded in the approved package information, with no invented discount or treatment guarantee. Review Forbidden Claims and Forbidden Selling Tactics under Guardrails. Our guide to medical tourism sales objections covers when a coordinator call can help resolve the decision.

Run one scenario, inspect it, then run the set

The runner supports an individual run from a scenario row and a batch run across all scenarios. Start with one case while refining its setup. Open the results drawer and review the score, turn count and message log. Once the case tests the intended behavior, run the set together to check whether a configuration change has affected another path.

  • Read the score against the rubric. It reflects the configured evaluation. Keep a separate record of blocking failures, such as booking despite a required missing document.
  • Use turn count to investigate friction. Read whether the conversation repeated a request, lost context or needed clarification. Fewer turns alone do not establish better handling.
  • Read the complete message log. Find the turn where the patient relationship, prerequisite, language or budget concern changed. Check the response against the expected outcome.
  • Verify claimed actions separately. The documented scenario results are the score, turn count and message log. If a run says it updated a record or completed an action, verify that outcome only where the product provides supporting evidence. If that evidence is unavailable, record the outcome as unverified.

For conversation investigation, Chats provides recorded lead messages, tool execution logs and captured runtime prompts. It is a separate review surface from the scenario results drawer. Use it when examining a controlled connected-line conversation or a live failure; do not assume every synthetic run appears there.

Correct the cause and repeat the affected tests

An outdated package answer calls for checking the approved source. A booking that bypasses an identity requirement calls for checking the configured prerequisite. Repeated document requests call for reviewing the request policy and conversation evidence. Change the relevant source or setting, then repeat the failed case and neighboring cases. Our post on keeping medical tourism answers up to date covers source maintenance.

Keep a review record of the scenario, expected outcome, configuration reviewed, actual result, evidence and person responsible for any fix. Repeat relevant tests after changes to knowledge, instructions, models, permissions or policies. Add representative failures from real enquiries to the next review set.

Publication is a separate decision. Save stores draft changes; Create Version publishes the configuration live. Before publication, resolve the blocking failures and identify any checks the simulation did not cover. Use a contact your team controls for actual WhatsApp delivery and human handover testing.

A passing score has a defined scope

Test results provide evidence for the scenarios covered. They do not guarantee future accuracy, establish clinical suitability or transfer the agency’s legal responsibility. Keep medical decisions with the appropriate clinician and continue reviewing exceptions after launch.

The AriaBee medical tourism operations overview shows the wider work across leads, documents, quotations, travel and finance. For this review, the manager needs concrete evidence: the right patient identified, prerequisites respected, files accounted for and the next action accurately recorded.

Frequently asked questions

How should a medical tourism agency test WhatsApp automation before launch?

Define expected outcomes, run controlled conversations and inspect the replies. Verify claimed actions only where the product provides supporting evidence; otherwise, mark them as unverified. Include a relative enquiring for a patient, missing required identification, medical-file uploads, a language change and a budget objection.

Where do I run automated scenarios in AriaBee?

Open AI Sales Agents, select the relevant agent and open Test. Configure a synthetic lead scenario, run it individually or run the full set, then inspect the score, turn count and message log.

Can a scenario test send real WhatsApp messages?

The global test configuration includes a permission for live outbound WhatsApp messages. Verify that it is off for simulation; the help guide does not state its default. Enable it only for a deliberate controlled delivery test.

What can I configure for a synthetic lead?

The documented scenario setup includes client name, language, opening message, personality parameters, treatment goals, budget concerns, stop conditions and test medical-file uploads. Global test configuration lets you choose the evaluator and default synthetic lead models, rubric criteria and severity levels.

Does a high scenario score prove that an action succeeded?

No. The documented scenario results are the score, turn count and message log. A reply claiming that an action completed is not sufficient evidence. Verify the action only where the product provides supporting evidence; otherwise, mark the outcome as unverified.

When should we repeat the tests?

Repeat relevant cases after changing knowledge, instructions, models, tools, permissions or agency policies. Re-run failed and related cases after a correction, and add representative live failures. Passing results cover the tested scenarios; they do not guarantee future accuracy or replace medical judgment.

Review your agency’s patient enquiry scenarios

Bring an example intake and your document requirements. Review how to configure scenarios, control outbound messaging and inspect the results in AriaBee.

How to test a WhatsApp chatbot for medical tourism