You want a conversational application for a chosen channel.
AI chatbot development →WhatsApp AI chatbot.
Support WhatsApp conversations with AI and your own business information.
Discuss your projectA chatbot or a customer service solution?
You want to improve how your department handles customer enquiries.
AI customer service →An assistant must also perform authorised actions in connected systems.
AI customer service agent →How does a WhatsApp AI chatbot work for your business?
A WhatsApp AI chatbot supports business conversations through a suitable WhatsApp Business channel. It can interpret a customer question, use agreed information sources and hand the conversation to your support team. I start with the intended customer journey, then examine account access, current channel terms and connections to your customer or ticketing system. We then design a bounded flow, such as practical product questions or collecting information for an employee. Further actions in other systems also require an explicitly designed software integration.
When it is a good fit
This fits businesses whose customers already contact them through WhatsApp and where an identifiable group of questions can be supported by maintained business information. A person or support team needs to own that information and handle escalations.
When to take a different approach
Feasibility depends on your business account, provider and permitted application. A general AI assistant and a business-support flow can be subject to different terms. Without an available escalation route, suitable information or confirmed access, I would not start a full bot rollout.
A WhatsApp agent needs reliable message processing
For a WhatsApp chatbot, receiving, understanding and replying are designed as separate steps. Messages and status updates may arrive at different times. We retain a recognizable message identity and check prior processing before creating another reply or CRM change. [1]
The conversation task remains bounded: which questions are answered, when a person takes ownership and which actions may be proposed? Platform terms, account capabilities and message types are checked at project start. A model must not assume permission to contact a customer or commit the business. The current WhatsApp policy uses a 24-hour customer service window after the last user message; approved templates apply outside it. Automation must provide a clear escalation route. [2][3]
Questions to settle first
- How do we prevent two replies to a repeated message?
- Who handles a conversation when the bot pauses or hands off?
Make acceptance measurable
- A replay and failure test checking for duplicate replies.
- A handoff in which bot and human do not respond concurrently.
Compare your options
Choose based on the task, the people involved and what you can maintain.
| Approach | Useful when | What to consider |
|---|---|---|
| Manual business inbox | Conversation volume is limited or most questions need personal judgement. | Little development work; employees still do most of the handling. |
| Existing support platform with WhatsApp | You need shared handling and the available product features fit the task. | Check channel support, user licences, AI options and available data connections. |
| Fixed WhatsApp flow | Conversations follow known choices or collect a few defined fields. | Straightforward to bound; unexpected questions need a recognisable support route. |
| Custom AI support flow | Variable questions need your own knowledge and existing support process. | Requires evaluation, account administration, integrations and maintenance of information and channel requirements. |
What I need to start
- Confirmed administrative access to the business account, number and selected channel provider.
- An assessment of the intended application against current account and regional terms.
- Customer-question examples, approved information and topics reserved for an employee.
- An available escalation route and handling agreements during and outside business hours.
- Agreements about customer identification, required permission, message templates and conversation storage.
How to assess the result
- Incoming messages are recognised and associated with the correct conversation or ticket.
- The bot uses agreed knowledge and makes no unauthorised commitments when information is missing.
- Customers can reach the agreed support route and relevant conversation context is handed over.
- The design handles messaging windows, permitted templates and opt-outs under current channel rules.
- Failed delivery, unavailable integrations and repeated messages are tested without duplicate actions.
What determines the cost?
A useful estimate starts with the scope. I distinguish development work from recurring usage and operating costs. These are the main drivers:
- The channel provider, existing support platform and required user licences.
- The number of flows, languages, knowledge sources and customer or ticketing integrations.
- Messaging at current rates, plus model usage and any hosting.
- Knowledge maintenance, conversation evaluation, account changes and human follow-up.
Your challenge.
A practical application.
Customers ask practical questions on WhatsApp that are answered manually each time.
A WhatsApp conversation may need help from a person; an AI customer service agent defines the tasks and handoffs involved. Through CRM integration, relevant contact information can be retrieved or recorded within agreed permissions and the capabilities of the selected WhatsApp solution.
What account access and channel requirements apply to the intended workflow?
Agree the first step
Use the prerequisites, acceptance checks and cost factors above to define your assignment. Agree who can approve the outcome and who will maintain the solution. The project approach explains discovery, delivery and handover.
Read the project approachYour questions.
Clear answers.
Can a WhatsApp chatbot always send free-form replies?
The current WhatsApp Business Messaging Policy permits free-form replies within the 24-hour window following a user message. Approved templates are required outside that window. Automated replies also require a clear, direct escalation route. We check how the current requirements apply to your setup.
Can the bot use our existing business number?
I investigate this against your current account setup and selected provider. Administrative access, available connection options and effects on your existing inbox are checked in advance. Migration or retention of every feature should not be assumed.
Can the bot look up an order or customer status?
Only if a suitable integration exists and customer access to that information is defined. A phone number or conversation name alone is not an authorisation design. We limit the data shown and decide which actions require employee approval.
Sources and technical basis
These primary sources explain the technical principles behind this guide. The project recommendations are OXIAE’s interpretation; they do not imply endorsement by the publishers. Product documentation applies to the named platform.
- Retry patternMicrosoft Azure Architecture Center
Considerations around retries and unintended side effects.
Link and relevance checked: - LLM06:2025 Excessive AgencyOWASP GenAI Security Project
Limiting agent communication and system actions.
Link and relevance checked: - WhatsApp Business Messaging PolicyWhatsApp / Meta
Current platform requirements for messages, templates and escalation.
Link and relevance checked:
Prepare your project conversation.
Exploring WhatsApp AI chatbot? Use the project checklist to describe your task, systems and intended outcome. Collect open questions and download a brief for your conversation.
Create your project brief