Conversational setup helps you create an AI agent by describing what you want it to do in your own words. AgentFlow uses your description and available business information to understand your use case, ask follow-up questions when needed, and generate an initial playbook for your agent.

AgentFlow automatically generates a persona for your agent. Procedures are generated only when your use case includes scenarios that require structured steps or workflows. For simpler use cases, such as answering frequently asked questions, the generated playbook may not include any procedures.

Creating your agent is the first part of the setup journey:

1. Create your agent with conversational setup.
2. Configure how your agent should work.
3. Test your agent before deploying it to your messaging channels.

In this article, you’ll learn how to create an AI agent using conversational setup and review the configuration AgentFlow generates for you.

Before you begin

To use conversational setup, you need:

• Access to the SleekFlow AI module

Before you start, prepare a clear idea of what you want the agent to do. It can also help to have relevant business information ready, such as:

• The products or services the agent should discuss
• Common customer questions
• Tasks or workflows the agent should handle
• Situations where the agent should assist or hand over a conversation

Access the conversational setup

Follow these steps to create an AI agent with conversational setup:

1. Click the  icon in the left navigation menu to go to the SleekFlow AI page
2. Click Create AI agent in the top-right corner.
3. You’ll be redirected to the AI agent setup page.
   

Describe your agent’s use case

Start by telling AgentFlow what you want your agent to do and the types of customer conversations it should handle.

You can get started in either of these ways:

• Select a suggested quick start.
• Describe your own use case.

Use a suggested quick start

AgentFlow may suggest use cases based on the business information available in your workspace.

Select a suggested quick start that matches what you want your agent to do. AgentFlow will use the selected use case as the starting point for conversational setup.

If none of the suggestions match what you need, you can describe your own use case instead.

If AgentFlow cannot generate suggestions from your available business information, the suggested quick starts will not appear.

Use your own use case

Enter a description of what you want your agent to do in the use case field.

Focus on the agent's purpose and the customer conversations it should handle. A useful description can include:

• Who the agent will speak with
• What you want the agent to help customers accomplish
• The types of questions or scenarios it should handle
• Important business processes it needs to support

For example:

• Answer product questions and recommend suitable products.
• Help customers check their order status and delivery timeline.
• Qualify new leads before handing them to the sales team.
• Answer appointment questions and collect the required customer details.

Be specific enough for AgentFlow to understand the outcome you want, but you do not need to write detailed agent instructions at this stage.

Avoid pasting a complete prompt or instruction set created for another AI agent into this field. Instead, describe what the agent should do and the types of conversations it should handle. You can refine detailed instructions in the agent's playbook after it is generated.

How AgentFlow uses your business information

When you submit your use case, AgentFlow uses available business information to better understand your request and generate an agent that fits your business.

Depending on the information available, AgentFlow may use:

• Your workspace's registered industry
• Your workspace's registered website URL
• Products and services described on your website
• Common questions or support information published on your website
•  
• Brand information and tone of voice available on your website

If some information is unavailable, AgentFlow continues with the information it can access and may ask you for additional details during conversational setup.

AgentFlow does not use previous customer conversations to generate your playbook.

Set up your agent with Copilot

After you submit your use case, AgentFlow’s Copilot starts setting up your agent.

Copilot uses your use case and available business information to begin generating the initial configuration, including the agent’s playbook. You can follow the setup progress on the canvas while AgentFlow works in the background.

Answer quick questions

While your agent is being set up, Copilot may ask a few quick questions to better understand how you want the agent to work.

For each question, you can:

• Select one of the suggested answers.
• Select “Other” and enter your own answer.
• Skip the question.
• Go back to edit an answer within the current round.

Click “Next” to continue to the next question.

If you skip a question, Copilot continues using the information already available.

Review and submit your answers

After you complete the quick questions, Copilot shows a summary of your answers for review.

Check that the information is correct before continuing.

From the review screen, you can:

• Click “Edit answers” to go back and update your responses.
• Click “Submit” to confirm your answers and continue the setup process.

After you submit your answers, Copilot uses them to refine the agent it is generating.

Review your generated agent

After generation is complete, review the setup AgentFlow created for you.

Check that:

• The agent’s purpose matches your intended use case.
• The generated rules and procedures are appropriate.
• The business information is accurate.
• Any assigned Knowledge is relevant.
• Nothing important is missing or incorrect.

If you want to make changes, you can edit the generated content directly or enter additional instructions in conversational setup.

Next steps

Your generated agent is a starting point. Before testing and deploying it, review and configure its playbook, Knowledge, Context & Response settings, integrations, and exit conditions.

See Configure your AI agent to continue setting up your agent.