This article covers the steps required to get a conversation bot up and running in messaging offered through the Web Widget on your website or help center. A conversation bot is a dynamic, customizable bot that helps guide customers to a resolution. Before you get started, make sure you've set up your messaging channel.
You can create a conversation bot in any style that suits your organization, from a simple greeting that welcomes customers before handing them over to a live agent to a complex conversation bot that gathers customer data, routes conversations to specific agents, proactively notifies customers of updates or changes, and more.
Prerequisite: Configure a basic messaging channel
If you plan to create a conversation bot, before you can do so, you’ll need to give it one or more channels where it can communicate with your customers. Messaging is the most commonly-used. You can offer messaging through a Web Widget for your website or help center or through mobile SDK for your Android or iOS app.
If you haven't already configured a basic messaging channel, see Getting started with messaging.
Understand the anatomy of a conversation bot
Your conversation bot’s behavior is determined by how (and whether) you configure its standard response, customized answers, steps, and intents.
The standard response
The standard response defines your bot’s basic behavior in response to the question a customer enters into the widget.
The standard bot response include:
- Start of conversation: Initial greeting when a messaging conversation is started.
- If the bot finds relevant articles: Response when a possible solution is found in your help center.
- After the bot answers: Response to customer feedback.
- If the bot needs clarification: Response when the bot can't determine an answer that matches a customer’s comment.
- If a question matches more than one intent (requires an intent model): Response to a comment that matches more than one intent.
- If the bot can't answer a question: Response when the customer's comment does not match any answer.
Answers
Steps
Steps are the parts of an answer that determine how an answer responds to customer input. You combine these steps into answer flows during the answer creation process in bot builder. When a customer’s query triggers that answer, the bot displays the behavior defined by those steps.
You can use any of the following step types in your answer:
- Send message: Displays a piece of text to the customer during a conversation.
- Present options: Displays up to 10 preconfigured quick replies the customer can interact with during a conversation.
- Show help center articles: Presents up to six help center articles to the customer during a conversation.
- Add carousel: Displays up to 10 informational panels the customer can scroll through, each with a link to an external URL.
- Ask for details: Collects information from an end user during the conversation.
- Ask if question resolved: Asks your end users whether an answer resolves their support issue and offers response options they can choose from.
- Make API call: Presents external data in your conversation or pushes data to an external endpoint.
- Transfer to agent: Escalates the conversation to an agent.
- Branch by condition: Adds conditional logic to an answer flow. You can use the logic to personalize a conversation based on an end user's data.
- Add business hours condition: Lets you use your schedule to branch the conversation.
- Link to another answer: Links one answer in a custom-built answer flow to another.
- Set variable: Used to set the value of an existing variable or of a new variable created within the step.
Intents
If you have an intent model assigned to your account, you can use intent in your answers. Intents are predetermined topics that represent customer questions and requests. They are drawn from a set of industry-specific topics developed by Zendesk using our anonymized customer service data.
You can assign intents to answers to help save time and increase bot accuracy. When you assign an intent to an answer, the bot knows to show this answer whenever it identifies the intent in a conversation with a customer.
You can connect intents with answers as follows:
- When creating an answer, you can apply suggested intents, allowing you to build more effective conversation bots.
- When viewing your intents, you can resolve intent suggestions that identify commonly asked but unanswered questions from customers to help you fill gaps in your bot’s knowledge.
Step 1: Create a conversation bot
Creating a conversation bot means you’re adding the ability to use the bot in a channel’s messaging configuration. As part of this process, you’ll configure branding and language settings for your bot.
To add a conversation bot
- In Admin Center, click Channels in the sidebar, then select AI agents and automation > AI agents.
- Click Manage conversation bots.
- Click Create bot at the top of the page.
Alternatively, you can clone an existing bot as a starting point for your new bot.
- In The basics, enter the information, then click Next.
- Enter a name for the bot. This name is displayed at the top of the widget where the bot is deployed.
- Add a custom avatar (optional). The custom avatar replaces the default bot avatar. The image can be a JPG, PNG, or GIF file and it must be smaller than 100KB. An image size of 50x50 pixels gives the best results.
- If you’re using multiple brands, use the drop-down to select a brand to apply to the bot.
- In Language, select a bot language from the drop-down for your bot then click Next.
- In Generative AI, select Turn on generative AI features if you want your bot to use generative replies and bot personas.
- Click Create bot at the bottom of the page.
Step 2: Configure your bot using a standard response and generative replies
Settings for the conversation bot’s standard response define the default responses to customer input in a live conversation.
As part of the response when the bot does not find an answer, you can enable the bot to use generative replies to create AI-generated responses based on articles in your help center.
To configure your bot’s standard response
- In Admin Center, click Channels in the sidebar, then select AI agents and automation > AI agents.
- Click Manage conversation bots.
- Click the name of the bot you want to edit.
- Click the Behavior tab, then click each message to edit it:
Each response is saved automatically as you edit it, but updates will not be presented to customers until you publish the bot.
Now that you've configured the standard bot response, you have a fully functional bot that you can publish and use without any other configuration. If you want to make your bot more powerful, you can create custom answers now, before you publish the bot, or you can come back and add answers to the bot at a later time.
Step 3: Design a customized bot using answers (optional)
Answers tell a conversation bot how to best respond to customer questions through the Web Widget or mobile SDK. An answer addresses a single question your customers might ask. You can build your own answer from scratch or use a pre-built answer template. If you have an intent model assigned to your account you can use intents to help create and improve your answers.
An answer consists of the steps the bot takes in response to customer input. Each step represents a message the bot can show your customer during a conversation that addresses a single question your customers might ask. Answers are entirely optional but are incredibly useful for adding personalization and complexity to your customer interactions.
To create an answer for your conversation bot
- In Admin Center, click Channels in the sidebar, then select AI agents and automation > AI agents.
- Click Manage conversation bots.
- Click the name of the bot you want to edit.
- Click the Answers tab, then click Create answer.
- Click Build your own answer, then click Next.
- Enter the Answer name, then click Next.
The name should be a brief, clear description of the issue addressed by the answer. For example, “Password reset,” “Request a refund,” or “Talk to a human."
- Enter Training phrases in the available fields, then click Add phrase if you need
to add more.
Training phrases teach the bot how to interpret questions so it can match the most relevant answer. You can add up to 100 additional phrases. See best practices for training phrases.
Alternatively, if you have an intent model assigned to your account, you might see up to three AI-powered suggested intents to use instead of training phrases. You can select one or more intents. If none of the intents work, click Train the bot yourself to add training phrases.
- Click Next.
The answer opens in bot builder where you can build out the bot’s response.
- Click Add step to create the first step in the answer.
Each answer needs at least one step only.
- Under Choose step in the configuration panel, select a step type, then enter options to
Configure the step.
For information on the available step types, see Understanding answer step types. Click Preview to view the step as a user.
- Add additional steps to the answer as needed.
For more information, see Adding and removing steps in an answer.
- Click Done when you’re finished.
The answer appears as Ready to publish status.
At this point, you can click Create answer to create another answer or you can click Publish bot to make your new answer available in the bot. Before new or edited answers are available to customers, you must publish the bot they’re attached to. When you publish the bot, all answers in the Ready to publish state are published.
Step 4: Publish your bot
Whether you are using a bot with the standard bot response or using a bot with custom answers, you’ll need to publish that bot before it appears in any of your messaging channels.
Publishing a bot to a channel means that any other messaging response – either the default messaging response or another bot – is removed from that channel and replaced with the newly published bot.
To publish a bot to a channel
- In Admin Center, click Channels in the sidebar, then select AI agents and automation > AI agents.
- Click Manage conversation bots.
- Click the bot you want to publish.
- Make any changes necessary to the flow, then click Publish bot in the upper-right corner.
- If the bot is not currently published to a channel, in Choose where your bot appears, select one or more channels, then click Publish to selected channels.
- If the bot is already live in a channel, the bot is updated in that channel.