Skip to main content

Documentation Index

Fetch the complete documentation index at: https://knowledge.goautonomous.io/llms.txt

Use this file to discover all available pages before exploring further.

​
Overview

Go Autonomous comes with a library of standard intents that cover common request types. But every business is different — you may receive request types that don’t map neatly to the defaults. Custom intents let you create new classifications and train the AI to recognize them using descriptions and real email examples.

​
Creating a custom intent

1

Choose where to place the intent

In Axon, hover between items in the intent tree to reveal the + Add intent line. Click it at the position where you want the new intent to appear. The category and subcategory are pre-populated based on where you clicked.
2

Enter the intent name

Give your intent a clear, descriptive user-facing name — this is what your team sees in Workstation and Flow. For example: “Urgent Reorder” or “Sample Request.”
3

Write a training description

Enter a training prompt that describes what this intent looks like. This description acts as a guide for the AI model. Be specific about what makes this type of request different from others.
4

Link training emails

Attach real email examples from Workstation as training data. The more representative examples you provide, the better the AI gets at recognizing this intent.
5

Save the intent

Click Save. The new intent appears in the tree with editor mode active. Save again to persist the metadata (creator name, creation date).
Good training descriptions are specific. Instead of “orders that are urgent,” try “requests where the customer explicitly mentions a deadline, rush shipping, or uses words like ‘urgent,’ ‘ASAP,’ or ‘expedite.’”

​
Training with email examples

Linking real emails to your custom intent is one of the most effective ways to improve classification accuracy. When you link an email:
  • The platform uses it as a positive example during classification.
  • Duplicate emails are automatically detected and prevented.
  • You can link emails from any request in Workstation — search by subject, sender, or content.

​
How many examples do you need?

There’s no hard minimum, but more examples generally improve accuracy. Start with 5–10 representative emails that clearly match the intent. Over time, add more examples — especially edge cases that the AI initially misclassifies.

​
Editing an existing intent

To edit a custom or standard intent:
1

Click the intent name

In the intent tree, click on the intent name (not the checkbox). This opens the train/rename modal.
2

Modify the name or description

Update the user-facing name or the training prompt as needed.
3

Add or remove training emails

Link additional email examples or remove ones that are no longer representative.
4

Save changes

Click Save. The “Last edited” metadata updates to reflect your changes.
If you modify a standard intent’s training data and later deactivate it, the custom training data is lost. The standard intent reverts to its original configuration if reactivated.

​
What’s next