# Three Common WhatsApp AI Chatbot Setups for Education Centres: Knowledge Base, Management System and AI Agent Education centres commonly use WhatsApp AI chatbots in three ways. The first is an **AI knowledge base**, which answers general questions from approved course information, fee schedules and centre policies. The second is an **AI connected to the education-centre management system**. After identity verification, it can look up a particular student's lessons, fee status, attendance and remaining lesson balance. The third is an **AI agent connected to the centre's systems**. In addition to reading data, it can use approved tools to submit a leave request, create a make-up request or start a follow-up workflow. All three can appear as a WhatsApp conversation, but they do not have the same data access or permission to act. This article introduces the three operating models directly through parent and student examples. It does not rank them. ## 1. AI knowledge base for general questions A knowledge-base bot answers from information supplied by the education centre. It can cover timetables, grade levels, fees, tutors, addresses, trial lessons, leave policies and bad-weather arrangements. ### Conversation examples - **Parent:** Which day is the Primary 5 English reading class? - **AI:** Wednesday from 5:00 pm to 6:30 pm. A Saturday morning class covers the same syllabus. - **Parent:** My child's English grammar is weak. Which class should we consider? - **AI:** The primary English consolidation programme covers grammar, reading comprehension and writing. Tell me the student's grade and I can list suitable classes. - **Parent:** How is the Form 1 mathematics class charged? - **AI:** It is charged by term. I can send the latest course and fee sheet. - **Student:** Which MTR exit is closest to the centre? - **AI:** The centre is a short walk from Mong Kok Station Exit E2. Here are the address and map link. - **Parent:** What happens if Signal No. 8 is issued before class? - **AI:** Face-to-face lessons are suspended while the warning is in force. The centre will send the day's arrangement through WhatsApp. - **Parent:** Is a make-up lesson always available after taking leave? - **AI:** Eligibility depends on the notice period and available places. The centre must check the student's record to confirm an individual case. - **Parent:** Can my child try a summer course before enrolling? - **AI:** Trial lessons are available for selected courses. Tell me the grade, subject and preferred time. A knowledge-base bot does not know who the person in the chat is. If a parent asks whether fees have been paid or when a daughter attends her next lesson, the bot should not guess. Identity verification and a system or staff lookup are required. ## 2. AI connected to the management system for read-only lookup After identity verification, this setup can read authorised student records and answer personalised questions. In this article it is deliberately **read-only**: it can look up data but cannot change records or carry out a request. ### Conversation examples - **Parent:** When is Chan Siu Ming's next English lesson? - **AI:** Saturday at 2:00 pm, taught by Miss Lee in Room 3. - **Parent:** Are both of my children attending lessons on the same day this week? - **AI:** Chan Siu Ming attends on Saturday at 2:00 pm; Chan Siu Mei attends on Sunday at 11:00 am. - **Parent:** I transferred the fee last week. Has it been received? - **AI:** The system shows the fee as received on 13 July with no outstanding balance. - **Parent:** How many mathematics lessons are left? - **AI:** Four lessons remain. The next lesson is Sunday at 11:00 am. - **Parent:** Was my son absent last week? - **AI:** The system shows him absent from Saturday's English class and present at Sunday's mathematics class. - **Parent:** Has my make-up request been confirmed? - **AI:** It is confirmed for 25 July at 4:30 pm in Room 2. - **Student:** Which workbook do I need, and when is the test? - **AI:** Bring English Reading Practice Book 3. The quiz is scheduled for 29 July. - **Parent:** Has a place opened in the Saturday class? - **AI:** The student remains on the waiting list and the status has not changed. ### What it can do - Verify identity before answering personalised questions. - Read current class, teacher, classroom, fee, attendance, make-up and waiting-list records. - Combine general policy with an authorised individual record. - Stop disclosure and hand over if verification fails or information is incomplete. ### What it cannot do - Submit leave, cancel a lesson or change a class. - Create or modify a make-up booking. - Alter student, parent, attendance or payment records. - Approve refunds, discounts or exceptions. - Access an unauthorised student's records. - Invent an answer when the system has no data. ## 3. AI agent connected to the system for controlled actions An AI agent can read authorised data and use configured tools to carry out a defined action within approved identity, permission and policy boundaries. Each action requires a real tool, a permission scope, workflow conditions and a human-handover rule. Read more: [AI Agent vs Chatbot: What Is the Difference?](https://oneflash.hk/en/blog/ai-agent-vs-chatbot-hong-kong) ### Conversation examples - **Parent:** Chan Siu Ming needs to miss Saturday's English class. - **AI Agent:** Please confirm that the leave date is 25 July. - **Parent:** Yes. - **AI Agent:** The leave request has been submitted. Reference: L-0725-018. - **Parent:** Which make-up times are available? - **AI Agent:** This request is eligible. Current options are Sunday at 3:00 pm or Tuesday at 5:30 pm. - **Parent:** Sunday at 3:00 pm. - **AI Agent:** The make-up request has been created and is pending confirmation. - **Parent:** I transferred the fee. Here is the receipt. - **AI Agent:** The receipt is attached to a payment-review item pending staff verification. I will not mark the fee as received until staff confirm it. - **Parent:** I want to change the notification phone number. - **AI Agent:** Complete identity verification first. A primary-guardian change may require staff confirmation. - **Parent:** I would like a Primary 4 English trial lesson on Saturday morning. - **AI Agent:** I have created a trial-lesson follow-up. Staff will confirm level and availability. - **Parent:** I missed the leave deadline for a special reason. Can a make-up still be approved? - **AI Agent:** This is outside standard policy. I will pass the information to staff but will not approve it. - **Parent:** I am unhappy with the last class-change arrangement. - **AI Agent:** I can record the details, create a follow-up case and route it to staff. The centre will decide the outcome and any remedy. ### What it can do - Submit leave, make-up, trial-lesson or follow-up requests within its permission scope. - Execute the next approved step in a configured workflow. - Create or update specifically authorised records. - Return a reference number, processing status and next step. - Notify designated staff and retain an action record. - Stop and hand over when it detects an exception or risk condition. ### What it cannot do - Independently approve refunds, exceptional discounts or fee waivers. - Decide a complaint or make an unauthorised promise. - Guarantee a class place, make-up place or teacher assignment. - Access or alter another student's data. - Delete important records or bypass centre rules. - Perform an action when the system has no corresponding tool or permission. - Replace staff judgement in sensitive or exceptional cases. ## How the three setups answer the same question For: “My son needs to miss his Saturday English class. Can he attend a make-up lesson?” - The **knowledge base** explains the general leave and make-up policy. - The **management-system-connected AI** verifies identity and looks up the student's class and records, but does not submit anything. - The **AI agent** checks eligibility, submits leave, lists eligible times and creates the selected request. Exceptions go to staff. The point is not that one setup is universally better. The result depends on the data and permissions behind the conversation. ## Basic controls for all three 1. **Identity verification:** disclose personal lesson, fee or attendance data only after verification. 2. **Permissions and records:** define readable data and executable actions, and retain records for important operations. 3. **Human approval:** staff decide refunds, discounts, complaints, sensitive-data cases and exceptions. Further reading: [Education Centre AI Agent System Guide](https://oneflash.hk/en/blog/education-centre-ai-agent-system-hong-kong), [AI Automation for Tuition Centres: 3 Do's and 3 Don'ts](https://oneflash.hk/en/blog/tuition-centre-ai-automation-dos-donts-hong-kong), and the [oneflash education-centre system](https://oneflash.hk/en/industries/education-centre-system). ## Frequently asked questions ### Can a knowledge base answer a student's personal lesson time? Generally, no. Identity verification and authorised system access are required. ### Does system access mean the chatbot can modify data? Not necessarily. Read-only access allows lookup without changing records. ### Can an AI agent submit leave or arrange a make-up lesson? It can help when tools, identity checks, conditions, permissions and handover rules are configured. ### What can be disclosed before identity verification? General courses, fees, addresses and policies—but not an individual's classes, attendance, payment or contact data. ### Can an agent act when the system has no matching function? No. It requires a real tool and permission. ### Does every setup need human handover? Yes, especially for incomplete data, failed verification, exceptions, refunds, discounts, complaints and sensitive decisions.