Using the Action Engine in Ask AI

Introduction

The Ask AI feature in your CRM has evolved from answering questions to performing actions. You can now use conversational commands to create contacts, update opportunities, reschedule appointments, and more—all without navigating menus or switching tabs. This is powered by an integrated action engine that interprets your instructions and executes them instantly.

Key Benefits

Using the action engine within Ask AI offers several advantages:

• Streamlined Workflows: Reduce multi-step processes to a single command.
• Contextual Actions: Retrieve data and act on it in the same conversation.
• Broad Capability: Supports actions across contacts, payments, automations, and other areas.
• Role-Based Security: Actions are limited by your user permissions.
• Audit and Confirmation: Destructive actions require confirmation, and all actions are logged.

How It Works

When you issue a command, the system follows these steps:

1. Parse Intent: Identifies the action (e.g., create, update) and the object (e.g., contact, opportunity).
2. Generate Payload: Maps your natural language to the required fields.
3. Permission Check: Verifies you have the rights to perform the action.
4. Confirmation: For destructive or bulk actions, it prompts for confirmation.
5. Execution: The action is performed, and results are displayed in the chat.

Command Syntax and Examples

You can use natural language to issue commands. Here are some examples:

• Create: "Create a contact named Sarah Lee with email sarah@example.com."
• Update: "Move opportunity #789 to the 'Closed' stage."
• Query: "Show me all appointments for next Monday."
• Delete: "Delete the coupon code WINTER2024." (requires confirmation)
• Bulk: "Tag all contacts from the January webinar with 'Webinar-Jan'."

Tip: Add "and show me the result" to your command to see a summary of the action.

Safety and Permissions

Your ability to perform actions via Ask AI is governed by your user role permissions. Destructive actions, such as deletions or bulk updates, require explicit confirmation. All actions are logged and can be reviewed in the audit logs. Each successful action command uses one AI credit; previews or denied commands do not use credits.

Best Practices

To get the most out of the action engine:

• Use clear and specific commands.
• Combine data retrieval with actions in a single thread.
• Leverage conditional logic for automated decisions.
• Review audit logs to understand what actions were taken.

Frequently Asked Questions

How can I view what was sent to the server?

Click the timestamp of the chat response to see the raw JSON payload in the audit log sidebar.

Can I undo an action?

You can use "Undo last action" within five minutes, or restore from backups if available.

Are there limits on bulk actions?

Yes, bulk actions are limited to 1,000 records per minute to maintain system performance.

Is this available on mobile?

Currently, actions are read-only on mobile; full functionality is planned for a future release.

Can I restrict certain actions?

Admins can configure permissions for actions like deletions per role under security settings.

What if the AI misunderstands my command?

The system will ask for clarification and will not perform any action until confirmed.

Next Steps

To get started:

1. Open Ask AI in your CRM and try a simple command like "Create a test contact named Example User."
2. Check the audit logs to review the action details.
3. Experiment with safe bulk commands to see how rate limiting works.
4. Review your AI controls to ensure they align with your security policies.