WhatsAppAI agents

Give every customer a useful answer.

Build WhatsApp agents that answer product questions, check orders and help customers take the next step. Connect the knowledge and business systems they need, with your team ready for the conversations that need a person.

Set up a WhatsApp agent Read the agent guide

Set up with your agent:

CursorCopied!

Ivory stoneware vase — Try the reply buttons. This example changes no real order.

Knowledge that belongs to your business

A useful answer starts with the right context.

Bring your product information, service policies and relevant customer details into the conversation. The assistant can explain an option using the same information your team would check.

Choose the sources

Connect the knowledge you want the assistant to use, and keep it current as products and policies change.

Recognize the customer

Use the conversation and the customer details they are authorized to access. Ask for what is missing before making a recommendation.

Business context for this conversationExample

Order #4821

Ivory stoneware vase. Ready for collection.

Collection policy

Customers can change their collection day before picking up an order.

Studio opening hours

Tomorrow · 10:00–18:00

One answer, informed by the relevant sources.

From conversation to action

Follow the task all the way through.

Connect the assistant to the system that owns the order, appointment or account. Keep the request, the customer's choice and the actual result together, so the conversation ends with a useful next step.

Collection changeExample

Customer request

“Can I collect my vase tomorrow instead?”

ConversationContinue the existing exchange

GoalChange collection day

An illustrated workflow. Your connected systems provide the knowledge, permissions and action results.

Customer care handoffExample

Mora customer care

Collection request · needs a person

“Could someone help me collect after closing?”

Customer needs: A collection option outside studio hours
Already checked: Order is ready. Normal hours do not work.
Next decision: Arrange an exception or suggest delivery

A person when it matters

Your team joins with the story intact.

Some requests need judgment, an exception or a reassuring human answer. Bring your team into the exchange with the customer's goal, the information already checked and the step that needs attention.

Define who handles these conversations and give the customer a clear way to ask for a person. Keep the context available when the next step is a call.

Explore customer conversations

Continue by voice with WhatsApp calling or explore Voice.

Start with one customer task.

Choose a question your team answers often. Connect its knowledge, rehearse the action and review the handoff before expanding the assistant's role.

Plan your WhatsApp agent

Using a coding agent to build the integration? See how MCP setup differs from a customer-facing WhatsApp agent.

Questions about WhatsApp AI agents

Is Meta Business Agent the same as Bird AI?

No. Meta Business Agent is Meta’s runtime connected to an eligible business number. Bird AI assistance is a distinct operating model; evaluate the knowledge, actions and handoff of the chosen setup.

Is MCP a customer-facing WhatsApp chatbot?

MCP lets an authorized agent or tool work with Bird resources. That developer integration does not by itself create a customer-facing assistant or define its behavior.

Can an agent change an order or booking?

Connect an authorized action to the system that owns the order or booking. Apply identity checks and customer confirmation where needed, then communicate the recorded result.

How do we judge whether an AI agent works?

Review task completion, accurate answers, failed actions and the quality of handoff. Include difficult and ambiguous requests in the evaluation.