Most WhatsApp chatbots work on fixed keywords. A customer types "refund" and gets a refund policy. That works for simple queries, but customers don't always talk in keywords. They say things like "I ordered something last week and haven't received it" or "Can I change my plan mid-month?"

AiSensy's AI Agent Builder is built for exactly these situations. It understands what your customer is actually asking, finds the right answer from your business knowledgebase (help center), and can even take real actions, like checking an order status, booking an appointment, or fetching a quote - all without writing a single line of code.

In this tutorial, we’ll walk you through how you can build your own AI Agent on WhatsApp using AiSensy.

By the end of this, you’ll have a fully functional AI Agent that can understand user queries, respond intelligently, and even take actions automatically.

Note: This is a big guide, so stock up on your coffee & snacks & let's begin!

Keep these two things ready:

Log in to your AiSensy account and click Agents in the left menu.

In the Agents Studio, Click Set up to start configuring your Chat Agent.

You will also see Manage AI Messages Plan. Use this option when you are ready to select or change the AI Messages pack used by your agent.

Note: The Chat Agent is currently the only available agent type. More agent types mwill be added in the future.

Now, give AiSensy your website URL so it can understand your business and prepare the first version of your AI Agent.

Enter the following details:

Once both details are added, click Train Agent.

AiSensy will read your website pages in the background and use that information to prepare your agent for testing.

As soon as training starts, your agent begins answering messages on the test number you just entered. You'll see a confirmation screen that says “Your agent is live,” along with a note that AiSensy is still reading your website in the background.

You don't need to wait for the crawl to finish. The agent's answers get sharper automatically as more of your website gets read.

From here you have two options at the bottom of the screen:

If you clicked Continue setup, you'll land on a screen called “What should your agent do?” This is where you turn on the pre-built capabilities you want your agent to have, right from the start.

Two are already checked for you: FAQ/Support and Human Handoff. You can leave them as is, add more from Add capabilities, or click Skip to move on without changing anything. Everything here is fully editable later, so don't overthink this screen.

The last screen in the wizard is titled “Anything special it should know?” This is where you can add Ground Rules, instructions your agent should always follow, no matter what the conversation is about.

Never promise same-day delivery. Do not invent prices, discounts, or product availability. Use only the information available in Knowledge.

This field is optional. Skip it if you want, you'll find the same Ground Rules field again inside the main dashboard, so nothing is lost if you leave it blank here.

Click Go to agent to land on the full Chat agent dashboard, this is where you'll spend most of your time setting things up.

The Chat agent dashboard is a single scrolling page with everything on it: Business Profile, Voice, Skills, Knowledge, Actions, and Settings, listed on the left as quick jump links.

A few things worth noticing at the top before you scroll down:

The Business Profile tells the AI Agent who it represents and gives it the basic context required to speak about your business.

The AI Chat agent will craft a description for your business under the "What your business does" section. Kindly note that the AI agent might take some time to go through website and generate this description.

Review the generated information carefully and edit anything that is incorrect, unclear or missing.

If you wish to craft it manually, don't worry - we have explained the instructions below for reference.

For example:

Keep this description factual. Do not add offers, prices or policies that may quickly become outdated unless you plan to update them regularly.

Enter the name the AI Agent should use while introducing itself.

The greeting is sent word-for-word as the first message when the AI Agent starts a conversation with your AI Agent. If you add one, write it exactly as you want the customer to receive it.

For example:

If you do not want a fixed opening message, you can leave this field empty.

This is the same Ground Rules field from the setup wizard, edit it here if you skipped it earlier or want to add more.

Add only rules that must apply across every conversation. Use Skills later for instructions that apply to one specific job, such as lead qualification or appointment booking.

The Voice section controls your AI Agent’s response style. It decides how the agent speaks to customers through its messages.

Choose the voice that best matches your business:

Use this field if you want to give the agent more specific instructions about how it should communicate.

For example:

You can change these instructions based on your business.

For example, a financial services company may ask the agent to avoid casual language, while a fashion brand may allow emojis and a more playful tone.

Next, select how detailed the agent’s replies should be:

If you are unsure, choose Medium. It gives customers enough information without making every reply unnecessarily long.

Skills define the specific jobs your AI Agent can perform.

Here is the simplest way to understand the difference:

For example, you can create Skills for:

Skills are optional. A basic AI Agent can still answer questions from the Knowledge Base without adding a custom Skill.

Click Add Capabilities to view the pre-built Skills available in AiSensy.

The available options currently include:

Use this Skill for general questions about your products, pricing, policies, business hours, locations, delivery areas or services. The answer is taken from the information added in Knowledge.

Use this Skill when:

Human Handoff should not be used simply because the agent has not tried to answer the question yet.

Click Add beside the capability you need, then click Done. Use the checkbox beside a Skill to enable or disable it.

Expand an enabled Skill by simply clicking on it to review and configure it.

Depending on the Skill, you may see the following options:

This explains the type of customer message that should activate the Skill. Pre-built Skills managed by AiSensy already include this logic.

Some Skills allow you to specify information the agent should collect before completing the job.

For example, before transferring a conversation to a human agent, you may ask the AI Agent to collect:

Enter the exact question the agent should ask and select the contact attribute where the answer should be saved.

For example:

Question: What is your order ID?

Attribute: Order ID

Enable Required if the agent must collect the answer before moving ahead.

Click Add Variable to add more questions.

Ask only for information that is genuinely required. Too many questions will make the conversation feel like a long form.

Note: Do NOT use the FAQ/Support skill for order-specific lookups, lead qualification, or returns, these need their own Skill or Action since they depend on live or customer-specific data. Use FAQ/Support only when the answer should come straight from Knowledge. For order-related queries, either use Human Handoff or build a Custom Skill (see below).

Connect an existing API or define a new one if the Skill needs live data or needs to complete an action.

For example:

An API is not required when the answer can come directly from your Knowledge Base.

You can take assistance from someone in your tech team to define a new API call or, if you wish to do it yourself, see the Understand and Configure Actions section below to see how to set up an API call.

If the job you need is not available as a pre-built capability, click Add Capabilities and select Create Custom.

You can start with a suggestion such as:

You can also describe the job in your own words.

For example:

Qualify customers who are interested in purchasing our WhatsApp marketing plan.

Click Draft with AI to let AiSensy AI agent generate the initial Skill details.

Then review the following fields:

Choose an icon and give the Skill a clear name, such as:

Explain which customer messages should activate the Skill.

For a Lead Qualification Skill:

Tell the agent how to complete the job in the correct order.

For example:

Review the AI-generated draft before clicking Add Skill. Make sure it matches your actual business process.

Knowledge is the business information your AI Agent uses to answer customer questions.

This is one of the most important parts of the setup. If your Knowledge is incomplete, unclear or outdated, the agent’s answers may also be incomplete or incorrect.

You can add information in three ways:

Use Crawl a Page to add information from a website or help article.

Paste the URL of the relevant page and follow the instructions shown on the screen. Add pages containing useful customer information, such as:

Do not add unrelated pages that will not help the agent answer customer questions.

Use Upload Files to add business documents such as:

Supported file types include PDF, DOC, DOCX, TXT and CSV. Each file can be up to 50 MB, and PDF files can contain up to 30 pages.

Use Paste Text to directly add information that is not available on a webpage or inside a document.

This is useful for short FAQs, internal policies, product notes or temporary information.

For example:

Orders are usually dispatched within 24 to 48 working hours. Customers receive the tracking link on WhatsApp once the order is shipped.

Actions allow the AI Agent to do more than answer questions. They help it update contact information, search Knowledge, transfer a conversation or call an external API.

You will see three types of Actions.

Business Actions appear when compatible HTTP actions or toolkits have already been created for your account. They allow the agent to use tools connected to your business.

Business Actions are pre-built connections — like Razorpay or Shopify — that you set up once under Integrations → App Store. Once connected, they show up here automatically so your agent can use them without any technical setup. If you haven't connected any tools yet, this section will say 'No actions yet.'

These are built into AiSensy and are always available to the agent:

You do not need to configure these built-in Actions separately.

Custom API Actions connect the AI Agent to your external systems.

For example, the agent can:

Click Add Custom API Action and complete the following fields.

Give the Action a clear name based on what it does.

For example:

Explain when the agent should use the API.

For example:

Use this Action after the customer provides an order ID and asks for the latest order status.

Group related API endpoints under one purpose, such as orders, appointments or crm. This helps the agent understand which related Action to use.

Sensitivity controls whether the agent should ask the customer for confirmation before running the Action.

The Low setting runs the Action without an additional confirmation message. Use it only for safe, read-only tasks such as checking an order status.

For actions that change data, make a booking, cancel something, initiate a refund or create a financial impact, use a setting that requires customer confirmation if available.

Add one to five worked examples to help the agent use the correct values while calling the API. This is especially useful when the API accepts several parameters.

Filled in: the action name, the reason to call it, and the auto-generated tool name shown right under “Intent shaping.”

If your developer gives you a cURL request, paste it into the box and click Parse & Fill. AiSensy will use it to fill the request method, endpoint and other available details.

You can configure:

Available authentication options include:

Use the exact method, URL, authentication and request values provided in the API documentation. Never expose private API keys or tokens in public documents or screenshots.

If you're building the request by hand instead, here's what each part means:

You'll see four options under the Auth tab: None, Bearer token, Basic auth, and API key (header). Think of these as different ways of showing your ID.

You don't need to know which one is “right.” Your developer will tell you the auth type their system uses. Your job is just to pick the matching option and paste in exactly what they give you.

Click Test Call to confirm that the request works before saving it.

Use safe test data provided by your developer. Avoid using a real customer’s sensitive information for testing.

Click Next: Variables after the test request is configured.

Variables are the values the AI Agent should collect from the conversation and fill at the time of the API call.

For an Order Status Action, variables may include:

For an Appointment Booking Action, variables may include:

If AiSensy does not detect any dynamic values, the request will be sent exactly as configured. Do not save a fixed customer ID, phone number, email address or order ID unless the API is intentionally supposed to use that same value every time.

Once the correct variables are confirmed, click Save Action.

Open the Settings section and enter a fallback message.

The fallback message is sent when the AI Agent cannot find a reliable answer or cannot complete the customer’s request.

For example:

A good fallback message should:

Before making the AI Agent available to all customers, add up to three test numbers.

Only these numbers can interact with the agent during the free trial.

Enter each WhatsApp number, click Save Numbers and make sure you have access to those numbers for testing.

AiSensy provides 500 free test messages which can only be used to test your WhatsApp AI agent on your testing numbers. The usage bar shows how many messages have been used.

Test messages are meant for building and improving the agent before purchasing an AI Messages plan.

Below the fallback message, you'll also find a fuller Human Handoff setup: a toggle, trigger phrases, a handoff message, and then a Handoff webhook URL, Webhook authentication, and Extra headers.

Most businesses can skip the webhook fields entirely. They only matter if you want a handoff to automatically notify a separate system, like a helpdesk or CRM, the moment it happens. If you just want your team to see handed-off chats inside AiSensy, fill in the Handoff message and leave the rest blank, nothing breaks.

Use the Train Your Agent panel to test conversations directly inside the builder.

Train mode answers using your latest unpublished edits. Use it while changing the Business Profile, Voice, Skills, Knowledge, Actions or Settings.

This lets you check changes before publishing them for customers.

Live mode answers using the currently published version of the agent.

Use it to compare the live experience with your unpublished draft changes.

Test the agent with the following types of messages:

Click New Chat when you want to restart the test without the context of the previous conversation.

Once the agent is answering correctly, publish your latest configuration.

The top bar shows the current status of the agent:

Publishing applies your latest draft changes to the agent. It does not automatically make a free-trial agent available to every WhatsApp customer.

After publishing, switch to Live inside the test panel and run a final check against the published version.

During the free trial, the AI Agent works only with the test numbers you added.

To allow the agent to answer messages from all your customers, select an AI Messages pack.

You can select your AI messaging pack by going to Flows 👉 Purchase messages 👉 Select your Message Pack 👉 Confirm your purchase

Available pack sizes include:

Note: You cannot go live with your AI agent on AiSensy's free forever plan. You need a paid plan to go live. AI agents is an add-on feature in AiSensy and you need to purchase it separate from the AiSensy subscription plan. For more info on plans, check the AiSensy app or you can also check our pricing page.
​

Pricing is shown inside your AiSensy account and may depend on your billing cycle, country and applicable taxes.

The AI Messages plan includes:

Complete the payment to make the agent available to all eligible WhatsApp conversations.

You can later use Manage AI Messages Plan from Agent Studio to review or change your pack.

Before allowing the agent to speak with all customers, confirm that:

Once these checks are complete, your WhatsApp AI Agent is ready to handle real customer conversations.

Your AI Agent comes with 500 free AI messages for testing on up to 3 test numbers. Purchase an AI Messages pack only when you're ready to go live with real customers.

No. You can build a Knowledge-based AI Agent, set its Voice, add Skills, test it and go live without coding. Technical help is only required when connecting a custom API.

No. Skills are optional. Add them when you want the agent to perform a clearly defined job, such as qualifying a lead or handling a wholesale enquiry.

No. Add your FAQs, website pages or documents in Knowledge. An API is only required when the agent needs live information or needs to complete an action in another system.

The classic Flow Builder works on exact keyword matching and fixed conversation paths. AI Agent Builder adds natural language understanding, Knowledge Base retrieval, live API Tool Calling, and dynamic routing, so your agent handles open-ended conversations instead of just scripted ones.

Yes. You can have some flows on the classic builder and others on AI Agent Builder within the same account.

Contact attributes are fields used to save information collected during a conversation, such as a customer’s order ID, budget, company name or preferred appointment date.

It uses the fallback behaviour configured in Settings. You can also enable Human Handoff so the conversation can be transferred to your team when required.

Train mode uses your unpublished draft changes. Live mode uses the version that is currently published.

Published means the latest configuration has been applied. You still need an active AI Messages plan before the agent can answer all customers.

Yes, you can easily pause and update your WhatsApp AI agent unlimited number of times.

You can add up to three test numbers and use the free test message allowance shown in your account.

Only values that should change with each customer conversation, such as order ID, phone number, email address or appointment date. Use the API documentation or ask your developer to confirm the correct variables. Do not enter random values.

Review customer conversations regularly. Add missing information to Knowledge, refine Ground Rules, improve Skill instructions and retest major changes in Train mode before publishing them.

Any response involving intelligence, like a Knowledge Base answer, a tool or API call, or the agent reasoning through a query, counts as one AI message from your quota.