Workflows Overview

Workflows let your AI agent take actions beyond just responding to questions. When triggered, workflows can send data to external systems, escalate to humans, or tag conversations.

What Are Workflows?

Workflows are automated actions that execute during or after conversations. For example:

• Customer asks to cancel → Workflow notifies your team
• Conversation ends → Workflow tags the topic
• Customer provides email → Workflow updates your CRM

Workflow Types

Webhook

Send data to any URL when triggered:

• Notify external systems
• Update databases
• Trigger other automations

Create webhook workflows →

Zapier

Connect to thousands of apps through Zapier:

• Create tasks in project management tools
• Add contacts to email lists
• Update spreadsheets

Make (Integromat)

Similar to Zapier, connect via Make:

• Complex multi-step automations
• Conditional logic
• Data transformation

Escalate

Hand off to human support:

• Send conversation to support team
• Include transcript and context
• Collect customer contact info

Tag

Automatically categorize conversations:

• Apply signal tags
• Track topics and intents
• Power analytics

Custom Reply

Send a specific response:

• Canned responses for specific triggers
• Rich content (links, buttons)
• Consistent messaging

When Workflows Run

Workflows are triggered by:

AI Decision

The AI decides to trigger based on:

• Customer intent
• Conversation context
• Specific requests

Intent Matching

You define phrases that trigger the workflow:

• "I want to cancel"
• "Close my account"
• "Stop my subscription"

Configure triggers →

Memory Conditions

Trigger based on customer memory:

• Customer has VIP status
• Customer had previous issue
• Customer is on enterprise plan

Workflow Flow

Customer Message
      ↓
  AI Processes
      ↓
Workflow Triggers? → No → Normal Response
      ↓ Yes
Execute Workflow
      ↓
Action Completes
      ↓
Continue Conversation

Use Cases

Collect Leads

When someone shows buying interest:

1. AI identifies sales inquiry
2. Webhook sends data to CRM
3. Sales team gets notified
4. AI continues helping

Process Requests

When someone requests an action:

1. Customer says "cancel my order"
2. Webhook calls your order API
3. API confirms cancellation
4. AI confirms to customer

Smart Escalation

When AI can't help:

1. AI attempts to answer
2. After threshold, offers escalation
3. Escalation workflow triggers
4. Support team receives notification

Automatic Tagging

At conversation end:

1. AI analyzes conversation
2. Tag workflow applies relevant tags
3. Analytics updated
4. Trends become visible

Workflow vs Custom Instructions

Getting Started

1. Identify actions you want automated
2. Choose the workflow type
3. Configure the trigger
4. Set up the action
5. Test thoroughly

Create your first workflow →

Managing Workflows

View and manage workflows at AI Agents > Workflows:

• See all workflows
• Enable/disable workflows
• View execution history
• Monitor performance

Best Practices

Start Simple

Begin with one or two workflows. Add more as you learn.

Test Thoroughly

Test workflows in preview before production. Check:

• Trigger conditions work
• Data sent correctly
• External systems receive data

Monitor Execution

Regularly check:

• Are workflows firing correctly?
• Are there errors?
• Do success rates look right?

Handle Failures

Plan for when workflows fail:

• Have fallback messages
• Monitor error rates
• Set up alerts if possible