Back to blog
Medical Tourism Operations

WhatsApp chatbot giving wrong answers? Keep medical tourism answers up to date

A transfer policy changes. Automated replies keep promising the old inclusion. Here is AriaBee’s workflow for editing approved sources, reviewing affected questions and checking promises already sent.

•Published: •7 minute read

If your WhatsApp chatbot is giving wrong answers about your medical tourism agency’s services, check the approved information behind the reply, correct outdated terms and retest patient questions. In AriaBee, staff edit the AI sales agent’s knowledge sources in Agent Studio’s web console. The AI sales agent’s next knowledge search uses the updated text, and questions that relied on the changed text move to Needs review for staff verification. Earlier promises still need a manual check.

Your agency already coordinates patients, clinic quotations and travel through WhatsApp. AriaBee is the AI operations employee for medical tourism agencies, commanded from WhatsApp. Owners and coordinators delegate work by text or voice; administrators use the web console to configure and review its behavior. Keeping approved service information current is one of those administrative jobs.

The transfer changed. The AI sales agent’s answer did not.

Consider this illustrative agency scenario. A hair transplant package previously included airport, hotel-to-clinic and departure transfers. The agency changes its standard offer: new bookings include arrival airport-to-hotel transport, while the other journeys are quoted separately. A patient asks on WhatsApp, “Are all my transfers included?” The AI sales agent’s reply repeats the old promise.

For the owner or operations manager, that sentence creates work immediately. Someone must check the quotation, establish what the patient was promised, speak to the transfer provider and resolve who pays. Absorbing the extra transport reduces the booking’s margin. Asking the patient for an unexpected payment creates a difficult conversation before treatment.

Adding another coordinator to catch these contradictions adds cost and training. The underlying problem remains: the agency changed its service, but the approved answer did not change with it. This is the same dependency we describe in medical tourism quotation management: package inclusions become commitments that travel coordination and finance must fulfill.

Find what the incorrect reply was relying on

Start with the exact patient question and answer. “The AI got transfers wrong” is too broad to investigate. “It promised departure transport to a new booking after that inclusion was removed” gives your team something specific to check.

  • The source still contains the old promise. Correct that source and check for other copies of the outdated inclusion.
  • The source says only “transfers included.” Replace the shorthand with the journeys covered, the relevant package and the effective date.
  • The source is clear and current, but the reply contradicts it. Inspect the conversation and tool execution logs, retest the question and have staff handle the unresolved answer.

Clear instructions and relevant knowledge sources are also among the fixes in Zapier’s incorrect-answer troubleshooting guidance. For this medical tourism example, checking the actual transfer wording gives you a concrete starting point.

AriaBee’s Agent Studio Chats tab lets staff inspect the AI sales agent’s messages and tool execution logs. The staff composer can pause automation and send a direct manual message while the team resolves the issue.

Update the approved source in Agent Studio

Our October 30, 2026 release added direct knowledge-source editing and re-ingestion in Agent Studio. This gives administrators a way to correct an existing source when an agency’s services change.

  1. In the web console, open AI Sales Agents and select the relevant AI sales agent.
  2. Choose Knowledge / RAG in the studio sidebar.
  3. Find the entry under Knowledge Sources. Use the eye icon to inspect it.
  4. Use the pencil icon to edit its title and text. Replace the outdated inclusion with the agency-approved terms.
  5. After the edit, the AI sales agent uses the new text at its next knowledge search.

The same Knowledge / RAG area supports imported files, webpages and text. If an old brochure and an FAQ both contain the previous transfer promise, inspect both. Editing one source is not a substitute for checking conflicting material.

Example wording for agency approval

“For new Standard package bookings from [effective date], arrival airport-to-hotel transfer is included. Hotel-to-clinic journeys and departure airport transport are quoted separately. For earlier bookings, staff must check the accepted quotation before confirming coverage. If the package or booking date is unclear, ask staff to verify.” This is illustrative policy text; replace it with your agency’s actual approved terms.

A clinic’s message to a coordinator’s private phone does not automatically update this source. AriaBee captures conversations on the connected official WhatsApp Business line; information from side channels must be routed through it. The knowledge-source correction described here still happens in the web console.

“Needs review” puts affected questions back in front of staff

After a source edit, questions that relied on the changed text move to Needs review. Open each question and confirm that the revised source still answers it, or write an answer. This is a staff verification step.

In the transfer example, review “Is airport pickup included?” alongside “Do you take me back to the airport?” The updated policy gives those questions different answers. A general “yes, transfers are included” would hide the distinction again.

Retest the questions patients actually send

Use these prompts as a small acceptance check for the illustrative policy. Decide the expected answer before running them:

  • “Does the Standard package include airport pickup?” — Check that the answer identifies arrival airport-to-hotel transport.
  • “Are all transfers included?” — Check that it distinguishes arrival transport from separately quoted journeys.
  • “Who takes me from the hotel to the clinic?” — Check that it does not promise free transport.
  • “Is my return trip to the airport covered?” — Check that departure transport is described according to the revised terms.
  • “I booked before the change. Do I keep my transfers?” — Check that staff review of the accepted quotation is requested.

Agent Studio’s Test tab supports scenarios with an opening message, language and treatment goals. Staff can run an individual scenario and inspect its message log. Include the original question that failed, then repeat relevant checks in the languages your team uses. Read the replies yourself; a scenario score alone does not establish that the package promise is correct.

Earlier WhatsApp promises need a separate review

Updating a source changes what the AI sales agent’s next knowledge search uses. Do not treat it as a correction of messages already sent or an automatic notification to affected patients.

Have the responsible coordinator manually review relevant conversations since the service change. Check the actual wording, accepted quotation and booking status. Record the agreed resolution and contact patients individually where clarification is needed. Prioritize patients whose arrival is approaching.

That review belongs with the people handling medical tourism patient coordination. A corrected FAQ helps future conversations; the coordinator still has to resolve an existing patient’s transport arrangements.

Make answer maintenance part of approving a service change: name the person who updates the source, the person who reviews affected questions and the person who checks earlier commitments. AriaBee provides the editing and review mechanics. Your agency supplies the approved terms and verifies the result. You can see the broader delegation workflow in our WhatsApp command showcase.

Frequently asked questions

Why is my WhatsApp chatbot giving wrong answers after our services changed?

Check whether its knowledge still contains the old service terms, whether another source conflicts with the update, or whether the wording is ambiguous. If the approved information is current and clear, inspect the failed conversation and test the exact question again.

Where do we edit AriaBee’s AI sales agent’s approved service information?

Use the web console: AI Sales Agents → select the AI sales agent → Knowledge / RAG → Knowledge Sources. The eye icon opens a source; the pencil lets you edit its title and text. This source-editing workflow happens in Agent Studio.

When does AriaBee’s AI sales agent use an edited knowledge source?

After an edit, the AI sales agent uses the updated text at its next knowledge search. Retest the affected questions to verify the resulting replies; an updated source does not guarantee every answer will be correct.

What does Needs review mean after a source changes?

Questions that relied on the changed text move to Needs review. Staff must open each one and confirm that the updated text still answers it, or write an answer. The status identifies work for staff verification.

Does editing a source fix promises already sent to patients?

Do not assume it does. Staff should manually review relevant conversations, compare promises with accepted quotations and resolve any differences with the affected patients. Source editing is not an automatic correction of earlier messages.

See how WhatsApp delegation works

See how staff command medical tourism operations through WhatsApp, with the web console available for configuration and review.

WhatsApp Chatbot Giving Wrong Answers? Fix Stale Replies