Conversation AI Flow Builder in ZipLeads lets you create inbound chat experiences using connected AI agents, actions and conversation paths. Instead of handling every task through one prompt, you can give specific jobs and tools to different agents. The visual canvas makes it easier to organise conversations, route requests and test how your bot responds. This article explains how to build, test and set up a flow-based bot.
What is Conversation AI Flow Builder?
Conversation AI Flow Builder is a visual canvas for creating inbound conversations using specialised AI agents and connected nodes.
Each AI Agent can have its own instructions and tools. You can connect agents through routes, collect contact details, transfer conversations to other bots, and set up actions based on what a contact needs.
For example, a clinic can build a bot that collects patient details, answers treatment questions, books appointments, and transfers billing questions to a separate bot.
Important: The new Flow Builder is designed for inbound conversations. Outbound campaigns and outbound opening messages are still available through supported prompt-based bot setups.
Key Benefits of Conversation AI Flow Builder
Flow Builder helps businesses organise complex conversations without putting every instruction in one prompt.
Specialised Agents: Give different tasks and instructions to individual AI Agents.
Visual Routing: Connect nodes to guide conversations through different paths.
Global Assistance: Use a Global AI Agent for requests that may come up at any stage.
Flexible Tools: Give agents access to the tools they need for their job.
Conversation Continuity: Use Wait and Bot Transfer to manage pauses and handoffs.
Easier Testing: Test conversation paths using a selected contact.
AI-Assisted Building: Create or change flows using plain-language instructions.
Understanding the Flow Builder Canvas
The canvas is a visual workspace where you can create, connect and set up conversation nodes.
New flow-based bots start with a basic structure containing Start Chat, an AI Agent and an End Node.
The main controls are:
Build: Add and set up conversation nodes.
Deploy: Set up inbound channels and Working Hours.
Test: Test the bot using a selected contact.
Add New Node: Add agents, tools and flow-control components.
Start Chat is the fixed entry point for incoming conversations. It runs when a chat begins and doesn't need a separate trigger.
Note: The new multi-agent canvas is different from the earlier Conversation AI Flow Builder screen. Existing prompt bots, form bots and older flow setups stay separate.
How to Build a Conversation AI Bot with Flow Builder
Creating a flow-based bot involves adding agents, connecting conversation paths, assigning tools, and testing the result before you turn it on.
Step 1: Create a Flow-Based Bot
Start by creating a bot in the Conversation AI workspace.
Go to AI Agents → Conversation AI.
Click Create Bot.
Choose Flow Based Builder.
Open the Build canvas.
Review the starting conversation structure.
Step 2: Add an AI Agent
AI Agent nodes handle specific parts of the conversation.
Click Add New Node.
Select AI Agent.
Give the agent a descriptive name.
Add instructions explaining what it's responsible for.
Connect the agent to the right node in the flow.
For example, name the agent Appointment Assistant and tell it to help contacts book appointments.
Step 3: Add Tools to the Agent
Tools help an agent complete its task.
Select the AI Agent node.
Click Add a Tool.
Choose the tool you need.
Complete any extra settings.
For an appointment agent, add the Appointment Booking tool and set up the required calendar settings.
Only add the tools that agent needs for its job.
Step 4: Add Conversation Routes
Routing sends contacts to the right part of the flow.
Add a Router where the conversation needs to branch.
Set up the route conditions or descriptions.
Connect each route to its destination node.
Review the connections.
Step 5: Add Additional Nodes
Additional nodes let you collect information and control how the conversation continues.
Depending on your workflow, you can add:
Sequential to perform set actions in order.
Wait to pause the flow.
Bot Transfer to hand the conversation to another bot.
Email Capture, Phone Capture or Text Input Capture to collect information.
Global AI Agent to handle requests across conversation paths.
End Node to mark the end of the flow.
Step 6: Review the Completed Flow
Reviewing the full canvas helps you spot missing routes or nodes connected the wrong way.
Check the connections between nodes.
Review agent instructions and assigned tools.
Confirm Router destinations.
Check what the Global AI Agent is responsible for.
Review Wait, transfer and ending paths.
Step 7: Test the Flow
Testing lets you check the experience before the bot goes live.
Click Test.
Select a contact for testing.
Start a test conversation.
Try different requests and routing scenarios.
Check agent responses, tool behaviour and handovers.
Adjust the flow if needed.
Important: Some tools can change contact records, appointments or other data in your account. Use a suitable test contact and review the actions performed during testing.
Step 8: Set Up Deploy and Activate the Bot
Deploy settings decide which inbound channels the bot can answer and when it's available.
Open Deploy.
Set up the Inbound channels.
Set Working Hours and any after-hours behaviour you need.
Review the deploy settings.
Open Bot Settings.
Turn the bot on using the bot activation control.
Important: The new Flow Builder is for inbound conversations. Outbound campaigns are separate, and the canvas has no dedicated Publish button.
AI Agent and Global AI Agent Nodes
AI Agent nodes decide how different parts of a conversation are handled.
AI Agent
An AI Agent handles a specific task using its own instructions and assigned tools.
For example, you can create separate agents to:
Qualify a lead.
Answer questions using a Knowledge Base.
Book an appointment.
An AI Agent handles the conversation when its connected path is active.
Global AI Agent
A Global AI Agent stays available throughout the conversation, not just within one branch.
It's useful for requests that can come up at any time, such as:
Asking to speak with a person.
Asking the bot to stop.
Asking for another supported action while a different agent is active.
Global AI Agents can also use their own assigned tools.
Tip: Use specialised agents for individual tasks, and Global AI Agents for requests that should stay available across the whole conversation.
Routing and Flow Control
Flow-control nodes decide how conversations move between agents and actions.
Node | Purpose |
Router | Sends conversations down different paths based on the conditions you set |
Sequential | Performs set actions in order |
Wait | Pauses the flow until a supported time or reply condition |
Bot Transfer | Hands the conversation to another bot |
End Node | Marks the end of a flow path |
Sticky Note | Adds internal notes to organise the canvas |
For example, a Router can send appointment requests to a booking agent and billing questions to a transfer path.
A Wait node can pause the flow for a set time or until a supported reply condition is met.
Tools and Actions in Flow Builder
Tools let an AI Agent do specific tasks beyond writing replies.
You can give tools only to the agents that need them, rather than making every tool available across the whole flow.
Supported tools include:
Tool | Purpose |
Knowledge Base | Answer questions using connected business information |
Web Search | Find information beyond the connected Knowledge Base |
Appointment Booking | Book appointments using your calendars |
Skills | Perform supported CRM tasks |
API Call | Interact with external systems |
MCP | Use tools from connected MCP servers |
Custom Message | Send set wording |
Email Capture | Collect an email address |
Phone Capture | Collect a phone number |
Text Input Capture | Collect information through a text response |
For example, a questions agent can use Knowledge Base, while an appointment agent can use Appointment Booking.
Important: Some tools need extra settings, permissions or external connections before you can use them.
Bot Transfer and Wait
Bot Transfer and Wait help conversations carry on when a different bot or more information is needed.
Bot Transfer hands a live conversation to another bot you've set up. For example, a clinic's front-desk bot can transfer a billing question to a billing bot.
Wait pauses the current flow until a set condition is met. Supported conditions include elapsed time, scheduled times, contact replies and team member replies.
Team replies can also resume certain reply-based waits.
These nodes help you avoid restarting a conversation when a contact comes back or a different bot needs to handle the request.
For more transfer settings, see Conversation AI – Transfer Bot Action.
Building a Flow with AI
Build with AI helps you create or change a conversation flow using plain-language instructions.
Instead of adding every node by hand, describe what you want the bot to do.
For example:
"Create a clinic front-desk bot that collects patient details, answers treatment questions, books appointments, and transfers billing questions to another bot."
Ask AI can draft a flow from your description or help you change the existing canvas.
Note: Build with AI is marked Beta. Review the agents, routes and tools it creates before testing or turning on the bot.
Testing and Deploying Your Flow
Testing helps you check that the bot follows the right conversation paths before it handles inbound messages.
Use Test to try the flow with a selected contact. Test common requests, alternative routes, booking actions and handovers.
The Deploy tab controls where and when the bot can respond:
Inbound: Set up supported inbound channels.
Working Hours: Set when the bot responds and any supported after-hours behaviour.
Managing channels and Working Hours use separate permissions.
Important: Setting up Deploy doesn't turn the bot on. You turn the bot on in Bot Settings. The new canvas doesn't have a separate Publish button or built-in version history.
Frequently Asked Questions
Q: What is the difference between Flow Builder and a prompt-based bot?
A: A prompt-based bot uses one main set of instructions. Flow Builder splits conversation tasks into connected agents and nodes.
Q: Do I need to set up a trigger for Start Chat?
A: No. Start Chat is the fixed entry point for an inbound conversation.
Q: Can different AI Agents use different tools?
A: Yes. Each AI Agent can have its own instructions and assigned tools.
Q: What is a Global AI Agent used for?
A: A Global AI Agent handles requests that may come up at any point in the conversation, such as asking for help from a person.
Q: Can Flow Builder transfer a conversation to another bot?
A: Yes. Use Bot Transfer to hand the active conversation to another bot you've set up.
Q: Can a flow wait for someone to reply?
A: Yes. Wait supports time-based and reply-based conditions.
Q: Can I create a flow using AI?
A: Yes. Build with AI can draft or change flows from plain-language instructions. The feature is marked Beta.
Q: How do I test a flow before turning it on?
A: Click Test, select a contact, and run sample conversations through your paths.
Q: Does clicking Deploy automatically turn the bot on?
A: No. Deploy sets up inbound channels and Working Hours. You turn the bot on in Bot Settings.
Q: Can Flow Builder run outbound campaigns?
A: No. This Flow Builder is designed for inbound conversations. Outbound campaigns are handled separately.