WhatsApp Chatbot: Create a Free AI Agent in 5 Minutes

Support teams that already get order, return, product, and account questions on WhatsApp can use an AI agent to handle much of the repeat work. The goal isn't to answer everything. It's to handle routine questions well and bring in a person when the issue needs judgment or investigation.

With Chatbase, you can build and test an AI agent on your business data, choose WhatsApp as a channel, and connect a WhatsApp Business number through Meta. A simple agent can be ready in about five minutes. Meta verification can add time when you connect the number.

Teams that already have an agent can move straight to the WhatsApp connection steps. From there, the setup comes down to preparing your WhatsApp Business account, connecting the number, and testing both the conversations the AI should handle and the ones it should hand off.

What are WhatsApp Chatbots?

A WhatsApp chatbot is software that replies to messages sent to a business on WhatsApp. You may also see the same type of automation described as a WhatsApp bot. It can answer common questions, collect details, guide customers through routine requests, or pass a conversation to a person.

Rule-based bots follow fixed paths. An AI WhatsApp chatbot can deal with more natural wording because it works from business knowledge and instructions instead of relying only on menus or exact keywords.

For example, a rule-based bot may recognize "return policy." An AI agent can also understand, "I bought this three weeks ago and haven't opened it. Can I send it back?" if the return policy is part of its approved knowledge.

For support teams, that can make WhatsApp useful for customer service automation, product questions, lead capture, and triage. The agent still needs clear limits for billing disputes, account changes, policy exceptions, and other cases that need human judgment.

WhatsApp Business App vs WhatsApp Business Platform

These two products serve different needs:

An AI agent that replies through a business number needs a WhatsApp Business Platform connection rather than only the Business App.

What do you need before creating a WhatsApp chatbot?

Have these ready before you connect the agent:

• A Facebook account that can complete Meta onboarding
• A Meta business profile
• A WhatsApp Business profile
• A phone number you can use for the integration
• An approved WhatsApp display name
• Business knowledge such as policies, product information, help-center content, or internal docs
• Instructions that define what the agent should answer and when it should escalate

The phone number is worth checking early. For the current Chatbase WhatsApp setup, a number connected to an agent can't stay active in the WhatsApp or WhatsApp Business app. A number already active there must have its WhatsApp account removed before the integration can use it.

If the number was previously connected through Meta Developer for business, you'll also need to turn off two-step verification before setup. Chatbase requires an approved display name before the number is connected.

1. Create a WhatsApp Business Account

You don't need to start the current Chatbase setup by choosing a separate WhatsApp Business Solution Provider and configuring provider webhooks.

The connection starts in Chatbase and then moves into Meta's onboarding flow. Before you begin, make sure your business details and phone number are ready.

During Meta onboarding, you'll choose or create a business profile, create or select a WhatsApp Business profile, fill in the required details, and add the number you want to connect.

Use a number you're comfortable dedicating to the integration. Don't use a personal WhatsApp number for testing unless you plan to remove its current WhatsApp account and use that number for the agent.

Businesses that already have the right Meta and WhatsApp profiles can select them during setup instead of creating new ones.

2. Choose and Set Up Your Chatbot Platform

Connecting WhatsApp won't fix weak answers. The agent still needs accurate information, clear instructions, and a handoff path.

You can build and test the underlying AI agent on Chatbase's Free plan. Connecting it to WhatsApp requires a plan with integrations, so the free plan is useful for getting the agent ready before the WhatsApp channel goes live.

Teams still comparing options can review WhatsApp chatbot platforms. Once the platform is chosen, the next job is getting the agent ready.

Create and train your AI agent

The current Chatbase onboarding flow is:

1. Click New AI agent.

2. Add the information the agent should learn from.

3. Choose where you plan to deploy it.

4. Select WhatsApp.

5. Continue to create the agent.

You can start with a website or add sources such as files, text, or Notion. For an initial support setup, start with a small set of clean, current information instead of loading the agent with old or mixed documents.

Start with the questions your team sees every day:

• Shipping and delivery rules
• Returns and refunds
• Product details
• Business hours
• Account or service policies
• Common troubleshooting steps
• Escalation instructions

The AI chatbot setup process goes deeper into training sources, instructions, testing, and configuration.

Set instructions and boundaries

Training data tells the agent what it knows. Instructions tell it how to behave.

Define:

• What the agent is allowed to help with
• Which sources it should rely on
• The tone it should use
• What it must not guess
• What details it needs before taking an action
• Which requests must go to a person

For support teams, the last point matters as much as answer quality. A bot that tries to resolve every case can create extra work if it gives a confident answer where a person should have stepped in.

Test before connecting WhatsApp

Use the Playground to ask questions that sound like real customer messages.

Don't test only exact FAQ wording. Try incomplete questions, follow-ups, misspellings, and requests that aren't covered by the knowledge base.

If your return policy says unopened items can be returned within 30 days, test both:

• "What's your return policy?"
• "I bought this three weeks ago and haven't opened it. Can I send it back?"

A weak answer to the second question is a sign to fix the source or instructions before connecting a live number.

Support teams can create a free Chatbase account and test the agent before paying for the WhatsApp integration.

3. Integrate Your Chatbot with WhatsApp

Yes, you can integrate a Chatbase AI agent with WhatsApp. Open the agent, go to Channels > WhatsApp > Setup, then complete the connection through Meta's onboarding flow.

Already have an agent? Sign in to Chatbase and open it before starting the connection.

The current Chatbase flow is:

1. Open the AI agent you want to deploy.

2. Click Channels in the left sidebar.

3. Find WhatsApp.

4. Click Setup.

5. Log in with your personal Facebook account.

6. Click Get started in Meta onboarding.

7. Choose an existing business profile or create one.

8. Create or select a WhatsApp Business profile.

9. Fill in the business profile information.

10. Add the phone number you want to connect.

11. Click Continue.

12. Wait for Meta to verify the information.

13. Click Finish.

14. Update the WhatsApp profile settings if needed.

The WhatsApp integration documentation has the current setup screens and phone-number requirements.

One point is easy to miss: choosing WhatsApp during initial agent onboarding doesn't connect a number. The actual connection happens later under Channels > WhatsApp > Setup.

After the connection is live, you can manage numbers under Channels > WhatsApp > Manage. Chatbase also lets you connect more than one WhatsApp number to an agent.

Collect contact details from WhatsApp

Chatbase can collect a user's phone number and name from WhatsApp when that information is available. Automatic collection is managed under Channels > WhatsApp > Manage > Leads.

For more specific lead-capture flows, ask only for the details the business needs. A quick WhatsApp conversation shouldn't turn into a long form split across ten messages.

Do you need to configure webhooks manually?

Not for the standard Chatbase setup above.

A developer building directly on the WhatsApp Business Platform may work with APIs and webhooks. With Chatbase, the normal setup starts Meta onboarding from the WhatsApp channel instead of asking you to copy provider credentials or webhook URLs by hand.

How do WhatsApp's 24-hour window, templates, and opt-in rules work?

Connecting the number doesn't remove WhatsApp's messaging rules.

The 24-hour customer service window

When a customer messages your business, WhatsApp opens a 24-hour customer service window. During that window, you can reply without an approved message template. Each new message from the customer resets the window.

Outside that window, you need an approved template to start the conversation again. WhatsApp explains the rule in its Business Messaging Policy.

When do you need WhatsApp message templates?

Templates are pre-approved messages used when you need to contact a customer outside the 24-hour window.

Chatbase supports Utility, Authentication, and Marketing templates. The WhatsApp template setup shows how to create them in WhatsApp Manager and use them from Chatbase.

For proactive messaging, outbound WhatsApp campaigns use pre-approved Meta templates. Chatbase currently includes outbound campaigns from the Standard plan.

Do customers need to opt in?

Yes. WhatsApp says a business can contact someone only when it has the person's mobile number and permission to send later WhatsApp messages or calls. Businesses also have to honor opt-out requests.

Build that permission into the checkout, signup, booking, or support flow where the customer gives you their number.

4. Test Your WhatsApp Chatbot

Once the number is connected, test through WhatsApp itself.

Use questions that match the work your team gets every day:

• "How long does delivery take?"
• "Can I return something I opened?"
• "Where is my order?"
• "Do you have this in another size?"
• "I was charged twice."
• "I need to speak to someone."

Then test the awkward cases. Leave out an order number. Ask for a policy exception. Give the agent a vague request. Ask something that isn't in the knowledge base.

The right result isn't always an answer. Sometimes the agent should ask a follow-up question. Sometimes it should stop and hand the conversation to a person.

Chatbase supports attachments in WhatsApp conversations, so teams that receive product photos, receipts, screenshots, or PDFs should test those cases before launch.

Test human handoff

WhatsApp allows automation during the customer service window, but its Business Messaging Policy also requires a clear route to human support.

An in-chat takeover keeps the customer in the same conversation instead of making them start over by email or phone.

With Chatbase Takeover, a support associate can step into an ongoing WhatsApp conversation. The AI stops replying, Chatbase creates a Helpdesk ticket with an AI-generated summary, and the associate takes over.

The AI can be turned back on after the human part of the case is finished. Chatbase currently includes Helpdesk from the Standard plan.

What should a WhatsApp chatbot not automate?

Some support issues shouldn't stay with the bot.

Send a conversation to a person when it needs judgment, identity checks, sensitive account access, or a policy exception. A wrong answer in these cases can create more work than the automation saves.

Typical examples include:

• Billing disputes that need investigation
• Account or security changes that require identity checks
• Refund or policy exceptions
• Complaints where the customer asks for a person
• Requests that depend on data the agent can't access
• Actions that fail or return conflicting information
• Cases with legal, safety, or privacy concerns

When the right integrations are configured, an agent can complete more tasks. That doesn't mean every task should stay automated.

A practical customer support workflow should define both the automated path and the human path. For teams using an AI customer support agent, the aim is to clear routine work without making harder cases harder.

5. Launch Your WhatsApp Chatbot

Before giving customers the number, run a final check:

• Confirm the WhatsApp Business profile is correct.
• Confirm the intended number is connected.
• Test the highest-volume questions.
• Test requests the AI should refuse or escalate.
• Check the handoff path.
• Review opt-in for proactive messages.
• Make sure required templates are approved.
• Tell the support team when to take over.
• Check that key knowledge sources are current.

A smaller rollout can expose gaps without putting the full queue at risk. Review the first real conversations, fix the obvious misses, then expand usage.

How does a WhatsApp chatbot fit into other support channels?

WhatsApp is often one part of a wider support setup.

A team may use WhatsApp for fast messaging, email for longer cases, live chat for website visitors, and phone for urgent or complex issues. The right mix depends on customer habits and the type of problem being solved. A broader customer support channel strategy can help define those roles.

For Meta channels, keep the knowledge and handoff rules consistent even though each channel has its own setup. A Facebook Messenger chatbot can handle Messenger conversations, while an Instagram chatbot covers Instagram messaging.

Teams that manage several social channels also need a consistent way to route conversations and hand them to people. That is where social media customer service becomes an operations issue, not just a bot setup task.

WhatsApp Chatbot for Business: Common Use Cases

Start with repeat requests customers already send through messaging.

Customer support

A WhatsApp agent can answer common questions about shipping, returns, policies, product setup, or troubleshooting. It can also collect useful context before a person joins.

Order and account questions

When the right systems are connected, the agent can help with requests that depend on live information, such as order status or account details.

For an online store, that can be part of a broader ecommerce chatbot setup across the website and messaging channels.

Product assistance

Customers can ask about product details, sizing, compatibility, or availability. The agent should answer from approved product data and avoid filling gaps with guesses.

Lead qualification

A chatbot can collect a few useful details before sending a sales-ready conversation to the right person. Keep it short. WhatsApp shouldn't feel like a form spread across a dozen messages.

Appointment and service requests

Service teams can use the agent to collect booking details, answer preparation questions, or trigger a workflow when the right action is configured.

How does Jumia use an AI agent on WhatsApp?

Jumia's J Force program uses Chatbase on WhatsApp to help independent agents find answers from its training material without waiting in a support queue.

According to the Jumia customer story, the deployment runs across eight African markets. Chatbase handles 50% of all support volume, 80% of queries are resolved without a human, and more than 1,500 conversations are handled per month.

When the AI can't resolve a request, Jumia sends the relevant claim form so the issue can move to human support.

Jumia didn't ask its agents to move to a new support channel. It put the AI in WhatsApp, where they already asked questions, and trained it on the material they needed.

6. Monitor and Optimize Your Chatbot

Real conversations show where the agent still needs work.

Review:

• Common questions
• Questions the agent can't answer
• Incorrect or incomplete replies
• Topics that trigger human takeover
• Failed actions
• Repeated clarification requests
• New questions missing from the knowledge base
• Customer feedback

Then fix the cause.

A missing policy belongs in the knowledge base. Repeated misunderstandings usually call for clearer instructions. Tasks that need live account or order data may belong in an action rather than static training content.

Don't optimize only for fewer human conversations. Some issues should reach people. What matters is that routine requests are handled correctly and difficult cases reach the team with enough context to act.

Can you create a free WhatsApp chatbot?

You can build and test the AI agent on Chatbase's Free plan, but the WhatsApp connection itself requires a plan with integrations. Chatbase's current pricing lists integrations from the Hobby plan, which starts at $40 per month.

Meta may also charge for WhatsApp Business Platform messages. Pricing is based on the recipient's market and the message category. Service messages are free, and utility messages sent in response to users are also free under the current model. Check WhatsApp Business Platform pricing for current rates.

So you can get the AI agent ready for free, but a live WhatsApp deployment may still have platform and messaging costs.

WhatsApp chatbot FAQs

Can I use my existing WhatsApp Business number with a Chatbase chatbot?

Yes, but the number can't stay active in the WhatsApp or WhatsApp Business app at the same time. Chatbase says the account tied to the number must be deleted from the app before that number is used by the agent. Check the current setup requirements before changing a production number.

Do I need a WhatsApp Business Solution Provider to use Chatbase?

You don't need to start the current Chatbase setup by choosing a separate BSP and adding webhook credentials. The standard flow begins in Channels > WhatsApp > Setup and then moves through Meta onboarding.

How much does a WhatsApp chatbot cost?

There are two costs to check: the chatbot platform and WhatsApp messaging. Chatbase plan costs depend on the features and usage you need. Meta pricing depends on the recipient's market and message category.

Can a WhatsApp chatbot hand a conversation to a human?

Yes. In Chatbase, a support associate can take over an ongoing WhatsApp conversation. AI replies stop, and the Helpdesk ticket includes an AI-generated summary so the associate has the context.

Start with real WhatsApp support questions

Start with the questions customers already ask on WhatsApp. Give the agent reliable information, make the handoff rules clear, and test the awkward cases before you put the number in front of your full customer base.

After launch, use real conversations to fix gaps in the knowledge and instructions. If you want to test the workflow first, build your WhatsApp AI agent for free and connect the channel when the agent is ready.