Workflow AI Builder

Workflow AI Builder lets you create complete automations by describing your goal in plain language. Instead of configuring triggers and actions manually, you can generate a workflow, review it, and refine it using AI. This guide explains access points, building steps, editing options, and best practices.

TABLE OF CONTENTS

• What is Workflow AI Builder?
• Key Benefits of Workflow AI Builder
• Entry Points
• How to Build Your First AI Workflow
• Configure Steps with Your Account Data
• Point and Edit for Precise Control
• Chat mode
• Best Practices for Effective Prompts
• Providing Feedback
• Beta Limitations
• Troubleshooting
• Quick Tips
• Frequently Asked Questions
• Related Articles

What is Workflow AI Builder?

Workflow AI Builder converts natural‑language prompts into end‑to‑end workflows. It accelerates automation setup and reduces manual configuration by generating triggers, actions, and structure you can immediately review and customize.

In Workflows, Workflow AI Builder generates a full workflow from a clear prompt. You can start from multiple entry points, then adjust the result using conversational edits, point‑and‑edit selection, or chat‑only brainstorming.

When your prompt names records that already exist in your account, such as a form, a pipeline and stage, an email snippet, a tag, a user, a workflow, or a custom field, Workflow AI Builder finds them and configures each step with them. The workflow is built and set up in one go, so you do not need to select these values manually afterwards.

What’s new in AI Builder v3 

AI Builder v3 streams results in real time and shows a step-by-step progress view while it builds or edits your workflow. Key improvements include:

• Live streaming progress cards during build and edit requests
  
  
• Multi-step changes in one request (for example, “rename, add a 24-hour wait, and turn off re-entry”)
  
  
• Session memory for follow-up edits (for example, “make that 48 hours”)
  
  
• Confirmation before starting fresh on a canvas that already has a workflow
  
  
• Session continuity when you close and reopen the AI panel
  
  
• Learn Mode improvements, including streaming answers and suggested follow-up questions
  
  
• White-label-safe output (AI avoids platform-branded terms in responses)
  
  
• Targeted bulk edits across all matching steps, a specified number of steps, or specifically identified actions and triggers, while leaving untargeted steps unchanged

Key Benefits of Workflow AI Builder

Understanding the value helps you decide when to use AI versus manual building and how to combine both for speed and control.

• Speed: Generate an entire workflow from a single prompt in seconds.
  
  Note: AI Builder now averages under 30 seconds for workflow generation (down from ~60 seconds), with no quality trade-off.
  
  

• Simplicity: Use everyday language—no need to remember specific triggers and actions.
  
  

• Consistency: Start with a structured draft, then iterate for accuracy.
  
  

• Flexibility: Edit with natural language or target specific steps visually.
  
  

• Scalability: Reuse prompt templates and best practices across teams.
  
  

• Configured with your data: Steps use the forms, pipelines, snippets, tags, users, workflows, and custom fields already in your account.

Entry Points

Knowing all the ways to open Workflow AI Builder saves clicks and helps you pick the fastest route based on your context. There are three ways to access the workflow AI builder:

Step1: From the Workflow List Page

Navigate to Automation > Workflows and click the Build using AI button. This opens a modal where you can enter your prompt.

You can also click on Create Workflow and Start from Scratch to go inside the Workflow Builder.

Step2: From the Prompt Box Inside Workflow Builder

When creating a new workflow from scratch, you'll find an AI prompt box in the workflow builder. Simply type your automation requirements there or you can also use voice dictation to describe your workflow.

Step 3: Using the AI Chatbot Assistant

Inside the workflow builder, you can also use the AI chatbot assistant to help build your workflow.

How to Build Your First AI Workflow

A structured path—prompt, generate, review, and refine—ensures the AI output aligns with your automation goals.

Step 1: Enter Your Prompt

Enter your prompt in any of the entry points mentioned above, or ask the chatbot to help build a workflow. You have two options to get started:

1. Write or Dictate your own custom prompt - Describe your automation needs in your own words

2. Use the provided templates - Select from pre-written prompt templates available in the interface.

Here are some example prompts you can use:

• "Send a welcome email series when someone fills out my contact form"
  
  
• "Create a birthday reminder workflow that sends SMS greetings."
  
  
• "Notify my team on Slack when a high-value opportunity is created."
  
  
• "Follow up with webinar attendees 24 hours after the event."
  
  

• "When the Video Production Request form is submitted, create an opportunity in the Marketing pipeline at the New Lead stage and send the Welcome Email snippet."
  
  

Optional: Include details like

• The workflow name
  
  
• Any workflow settings you want applied
  
  
• The tools/apps to use (for example, “create a task in ClickUp”)
  
  
• How steps should use data from earlier actions (for example, “use the GPT output in the email body”)
  
  

• The exact names of records in your account that steps should use, such as a form, pipeline and stage, email snippet, tag, user, workflow, or custom field
  
  

Workflow settings you can manage with AI In your prompt, you can ask AI Builder to view, enable, disable, or update common workflow settings, including:

- Allow Re-entry

- Allow multiple Opportunities

- Stop on Response

- Timezone and Time Window (business hours)

- Sender Details (From Name, From Email, From Number)

- Conversations – Mark as Read

- Workflow name (rename during generation or rename an existing workflow) Example prompt:

Rename this to Appointment Follow Up v2, update the sender email to support@mybusiness.com, restrict messages to Monday–Friday 9 AM–6 PM, and confirm re-entry is off.”

Note: The Sender Details option listed above is a workflow setting. Workflow AI Builder can also update From Name and From Email across all or selected email actions through a scoped edit instruction.

Step 2: Generate the Workflow

Click Build Workflow button or hit Send (depending on your entry point). You'll see a loader while AI processes your request and builds the workflow.

While the loader runs, workflow generation typically completes in under 30 seconds on average. If it takes longer, try simplifying the prompt and generating again.

Note: Autosave is available for Workflow AI. When autosave is enabled, your AI-generated workflows are automatically saved the moment they're created and every subsequent edit you make through AI saves in the background.

Step 3: Review and Customize

Once the AI completes building your workflow, you will get the summary of the workflow created as well apart from the automation:

• If you used entry point 1, you'll be taken to the workflow builder
  
  
• If you used entry points 2 or 3, the workflow will be generated directly in your current builder view
  
  

You can then:

• Review the overall workflow structure
  
  
• Check individual trigger configurations
  
  
• Verify action settings and parameters
  
  

• Confirm that each step uses the forms, pipelines, snippets, users, and fields you named
  
  

Step 4: Edit Using AI

After reviewing your workflow, refine it with natural-language prompts in the AI chatbot. Describe the change and define its scope. Specify whether the change applies to all matching steps, a set number, or particular actions and triggers. Workflow AI Builder updates only that scope and leaves untargeted steps unchanged

You can also conversationally edit complex actions like If/Else and Wait steps. This includes updating conditions, adding or removing branches, changing wait settings, and refining existing logic without leaving the workflow conversation.

You can also ask AI to:

• Rename the workflow
  
  
• Check, enable, or disable workflow settings (re-entry, multiple opportunities, stop on response, timezone/time windows, sender details, and mark conversations as read)
  
  
• Update steps so later actions reference outputs from earlier steps (for example, inserting AI output into a message)

Undo and Redo AI-Generated Changes

You can undo and redo changes made by the Workflow AI Builder during your current editing session.

• Press Ctrl+Z on Windows or Cmd+Z on Mac to undo the last AI-generated change and restore the previous workflow state.
  
  
• Press Ctrl+Y on Windows or Cmd+Y on Mac to redo the AI-generated change you just undid.
  
  

Undo and redo work with AI-generated changes such as adding actions, editing existing steps, or restructuring the workflow.

As you continue building with AI, previous AI-generated versions remain available in the undo history for the current session. You can also use Recent Changes in the Workflow Builder to review recent edits.

Configure Steps with Your Account Data

Workflow AI Builder finds existing account records named in your prompt and configures matching triggers and actions as it builds. It reuses existing tags, workflows, and snippets instead of creating duplicates. Use the exact record names as they appear in your account for the best match.

You can name any of these records in your prompt:

Step 1: Name the records in your prompt

List each step and the exact record it should use, as it appears in your account.

A prompt that names the form, pipeline, stage, snippet, tag, user, workflow, and custom field to use.

Step 2: Let AI Builder find them while it builds

As AI Builder builds the workflow, the progress view displays each matched account record, including Found form, Found pipeline, and Found email template. This lets you see which records AI Builder identified before reviewing the configured steps.

The progress card lists the form, pipeline, and email template found in the account.

Step 3: Review the configured steps

Open the trigger and each action to confirm the values. In this example:

• Form Submitted: The trigger filter uses the existing form named in your prompt.
  
  
• Create Opportunity: Selects the specified pipeline and stage from your account.
  
  
• Send Email: Links the existing email snippet, including its subject and body.
  
  
• Add Tag: Uses the existing tag instead of creating a duplicate.
  
  
• Create Task: Assigns the task to the named user.
  
  
• Add to Workflow: Selects the existing workflow named in your prompt.
  
  
• Update Contact Field: Selects the existing custom field.
  
  

Form Submitted trigger: the filter uses the existing form.

Create Opportunity: pipeline and stage from the account.

Send Email: linked to the existing email snippet.

Create Task: assigned to a user in the account.

Update Contact Field: the existing custom field is selected.

Note: Use the record name exactly as it appears in your account. If a step uses a different record than you intended, change it with a conversational edit or select the correct record in the action.

What You Can Edit

The conversational editing feature currently supports:

Actions:

• Add new actions anywhere in your workflow
  
  
• Remove unwanted actions (both single and multi-path)
  
  
• Replace existing actions with different ones
  
  
• Modify action configurations and settings
  
  
• Update actions to use values from previous steps (data linking across actions)
  
  
• Add or update actions that use integration apps (for example, task creation actions)
  
  
• Move actions to different positions in the workflow

Triggers:

• Add additional triggers
  
  
• Remove existing triggers
  
  
• Replace triggers with different types
  
  
• Modify trigger configurations

Complex Edits:

• Combine multiple changes in a single prompt.
  
  

Limitations:

• Combine multiple changes in a single prompt
  
  
• Edit If/Else actions conversationally by adding, removing, or updating segments within branches, changing AND/OR operators for branches and segments, and adding or removing branches. For complex requests, AI Builder may ask clarifying questions to confirm which branch or segment to update.
  
  
• Edit Wait actions conversationally by updating wait attributes, adjusting reply or window settings, adding or removing timeout branches, or changing the wait type.
  
  

Targeted and Bulk Edits:

Use one instruction to apply the same change across all matching steps, a specified number of steps, or particular actions and triggers. Workflow AI Builder updates only the scope you define.

Supported targeted edits include:

• Updating copy across multiple email, SMS, or supported communication actions
  
  
• Changing pipeline stages across all or specified actions and triggers
  
  
• Setting From Name and From Email across all or selected email actions
  
  
• Applying exact copy to a defined group of actions without changing the remaining actions
  
  

Example Edit Prompts 

• "Add a 3-day wait between the welcome email and the follow-up"
  
  
• "Replace the SMS action with a Slack notification"
  
  
• “Update the SMS copy in only the first two actions.”
  
  
• “Replace the copy in the first 20 email actions and leave the remaining email actions unchanged.”
  
  
• "Change the trigger from form submission to contact created."
  
  
• "Add an if/else condition to check if the contact has made a purchase."
  
  
• "Remove the timeout branch from this wait action."
  
  

Note: For multi-path deletion, the AI may ask clarifying questions to ensure the changes match your intentions.

Point and Edit for Precise Control

You can target workflow steps by defining the scope in your instruction or by using Point and Edit. Use Point and Edit when visually selecting actions is clearer than describing them in the instruction

How it works:

1. Click Point and Edit in the AI chatbot interface to activate selection mode.
   
   
   

2. Select Actions to Edit: click to select one, click multiple, or hold Shift and drag to select a range. Selected actions are highlighted.
   
   
   

3. Describe your change in the chat; AI applies it only to the selected steps.
   
   

   
   

Example Point and Edit Scenarios:

• Select 3 email actions in a sequence and say "Change all these to SMS messages"
  
  
• Select a specific task action and say "Change the text to be more urgent"
  
  
• Select multiple actions in a branch and say "Move all of these after the tag action"
  
  
• Select a group of actions and say "Delete these selected actions"
  
  

Benefits of Point and Edit:

• Precision: Edit exactly what you want without affecting other parts
  
  
• Visual Clarity: See what you're editing before making changes
  
  
• Efficiency: Make bulk changes to multiple selected actions at once
  
  
• Reduced Errors: Eliminates confusion in complex workflows with similar actions
  
  

Tip: Point and Edit is especially useful for workflows with 10+ actions or multiple branches where verbal descriptions alone might be ambiguous.

Clarifying Agent in Workflow AI Builder

The Clarifying Agent helps Workflow AI Builder create better results by asking for missing details before it builds a new workflow or applies changes to an existing one. This improves first-pass accuracy, reduces unnecessary revisions, and helps the AI use the right context before it acts.

When your request is too broad or leaves out important details, Workflow AI Builder may ask a short series of clarifying questions before proceeding. This helps the AI avoid making incorrect assumptions when building or editing a workflow.

The Clarifying Agent may appear when key details are missing, such as:

• The workflow trigger is not defined
  
  
• The message channel is not specified
  
  
• The timing is unclear
  
  
• An unsupported channel is requested
  
  

The questions are designed to be quick and easy to answer. Users can select from available options or type a custom response. If a user prefers not to answer, they can skip the question and let the AI decide.

Workflow AI Builder asks up to three clarifying questions before continuing.

When the Clarifying Agent appears

The Clarifying Agent appears when Workflow AI Builder detects that your request does not include enough detail to generate a reliable result. This applies to both new workflow requests and edit requests for existing workflows.

Examples include:

• “Follow up with the customer” without a trigger
  
  
• “Send a message sequence” without a channel
  
  
• “Remind them later” without timing details
  
  
• Requesting a channel that is not supported

How to answer clarifying questions

Clarifying questions help the AI Builder narrow down your intent before it proceeds. Answering them gives the AI more context and usually leads to a better result on the first attempt.

1. Review each question shown by the AI Builder.

2. Select one of the provided options, or type your own response.

3. Continue through the questions if more than one appears.

4. Skip any question if you want the AI to decide for you.

5. Let the AI Builder generate or edit the workflow using the added context.

Why this improves workflow generation

Clearer inputs lead to better outputs. By gathering missing details before generation or execution begins, the Clarifying Agent helps reduce revision cycles and improves the quality of workflow builds and edits.

Chat mode

Chat Mode enables planning without building, so you can co‑design requirements with AI before committing changes.

How to use Chat Mode

• Click Chat Mode in the AI panel to enable brainstorming mode.
  
  

• Brainstorm with AI to plan triggers, actions, branches, and timing.
  
  

• When ready, disable Chat Mode and ask AI to build the workflow.
  
  

Brainstorm with AI

• Brainstorm first: Decide what to build and make AI your brainstorming partner
  
  
• Create a plan: Go back and forth and create a plan for the actions and triggers
  
  

Ask AI to build it

• Once finalized, disable chat mode
  
  
• Then ask AI to create it
  
  

Post-generation To-Do List (Complete Before Publishing)

After AI generates the workflow, review the To-Do List (Complete these steps before executing your workflow) for any items that still need your input, such as credentials, required fields, or custom values.

Select a To-Do item to jump directly to the related action, then complete the missing details before publishing. This helps prevent incomplete workflows from going live.

Common items in the To-Do List may include:

• Connecting integration credentials
  
  
• Selecting the correct account
  
  
• Completing any action settings that AI Builder could not resolve automatically
  
  
• Adding required field values
  
  
• Defining step-specific settings when multiple options are available
  
  

Best Practices for Effective Prompts

Clear prompts improve accuracy. Provide specifics on timing, channels, conditions, and content so the AI builds what you expect.

• Be specific and clear. Example: “When a contact books an appointment, wait 24 hours, then send a confirmation email and a reminder SMS one day before the appointment.”
  
  

• Include key details. Timing, channels, conditions/filters, and content type.
  
  

• Use action verbs. Start with Send, Notify, Create, Update, Wait, Check if.
  
  

• Define the edit scope. State whether the change applies to all matching steps, a specific number of steps, or particular actions and triggers.
  
  

• Name the exact records. Use the names of forms, pipelines, stages, snippets, tags, users, workflows, and custom fields exactly as they appear in your account.

Providing Feedback

Feedback helps improve generation quality and informs future enhancements.

• A feedback chatbot opens when a workflow is created.
  
  

• Click thumbs up if the workflow meets expectations, or thumbs down to provide improvement details.
  
  

• Share whether the automation matched your intent, what was missing/incorrect, and suggestions.

Beta Limitations

Knowing today’s limits helps you plan validation steps before activating a generated workflow.

• Manual review required: Verify triggers, actions, and configurations before publishing.
  
  

• Configuration verification: Some complex configurations may require manual adjustments.
  
  

• Testing: AI cannot test workflows. Perform manual tests.

Troubleshooting

If the first draft misses the mark, iterate on the prompt and narrow the scope to guide the AI.

• Refine your prompt: Add detail or split complex goals into smaller parts.
  
  

• Wrong trigger: Specify the exact trigger type in your prompt or edit using AI.
  
  

• Missing actions: List desired actions explicitly or add via conversational edits.
  
  

• Incorrect timing: Use precise time references (e.g., “after 2 days”).
  
  

• Wrong record selected: Use the exact name of the form, pipeline, snippet, user, or field as it appears in your account, or fix it with a conversational edit.

Quick Tips

These tips reduce setup time and prevent common mistakes when adopting AI‑assisted building.

• Start with simple workflows to learn capabilities; then increase complexity.
  
  

• Always test before publishing.
  
  

• Provide feedback to improve results over time.
  
  

• Revisit periodically for new features and improvements.

If a generated workflow is not ready to publish, check the Post-Generation To-Do List in the AI panel. It highlights unresolved actions and triggers that still need manual input.

Frequently Asked Questions

Q: Can I edit the generated workflow using AI?

Yes. Use conversational edits to add, remove, replace, move, or modify actions and triggers. You can also edit multi-path configurations for If/Else and Wait actions through conversation.

Q: Can I target only certain steps when editing?

Yes. Define the scope in your instruction by identifying all matching steps, a set number, or particular actions and triggers. Workflow AI Builder applies the change only to that scope. You can also use Point and Edit to select one or more workflow actions visually.

Q: What if the workflow doesn’t match my intent?

Refine your prompt with more detail or perform conversational edits. You can also plan in Chat Mode before building.

Q: Do I still need to test?

Yes. Always validate triggers, actions, timing, and outcomes with a live test before activating.

Q: How do I disable Workflow AI Builder?

You can disable it at the agency or Sub-Account level.

Agency: Go to Settings → Labs and disable Workflow AI Builder.

Sub-Account: Go to Automations → Global Workflow Settings → Workflow AI and toggle off AI Builder.

Q: Why do items appear in the Post-Generation To-Do List?

Some settings still require your input, such as integration credentials, account selection, or unresolved action settings. Complete these items before publishing.

Q: Is an AI-generated workflow always ready to publish right away?

Not always. Review the Post-Generation To-Do List after generation and complete any required fields before you publish the workflow.

Q: Can Workflow AI Builder use forms, pipelines, and other records that already exist in my account?

Yes. Name them in your prompt exactly as they appear in your account. AI Builder finds each one and configures the matching step, for example the form filter on the trigger, the pipeline and stage on Create Opportunity, the linked snippet on Send Email, the assignee on Create Task, and the field on Update Contact Field.

Q: What if a step uses the wrong record?

Ask AI Builder to change it with a conversational edit, or open the action and select the correct record. Always review each step before publishing.

Related Articles

• Workflow Builder: Learn More Using AI

• Workflows Landing Page

• AI-Powered Email Generation in Workflow Action

• Highlighting & Resolving Errors in a Workflow