Using AI-Powered Card Suggestions

Guru’s AI Suggest feature uses machine learning (ML) to proactively surface relevant Guru Cards in web-based support and email tools. It learns from your team’s interactions over time, refining suggestions to match specific workflows—saving time and reducing search effort.

👥

Access Required

  • Available on the Enterprise plan.
  • Up to 5 relevant Cards appear in the Guru browser extension over supported applications including: Drift, Freshchat, Freshdesk, Gmail, GoSquared, Intercom, Kustomer, LiveChat (Chat/Tickets), LivePerson, Salesforce Live Agent, Salesforce Service Cloud, SnapEngage, Zendesk (Chat/Tickets).

Enabling AI Suggest

  1. Navigate to Manage > Apps and Integrations > AI Suggested Text.
  2. Toggle ON for each third-party app where you want Guru to suggest Cards.

Accessing suggested Cards

You can access suggestions in several ways:

  • Guru browser extension open: A yellow dot appears next to the ⚡️ icon. Click it to view suggestions.
  • Guru browser extension collapsed: The side tab on your browser turns yellow. Click to open the suggestions list.
  • In supported apps: Click the Guru button. If suggestions exist, the button will show how many Cards (up to 5) are available.

✍️

Note

The yellow dot also appears when a Knowledge Trigger is activated by matching conditions on the page.

Helping Guru improve its suggestions

If suggestions aren't helpful, continued use of Guru’s browser extension will improve results. Interactions like viewing, copying, and engaging with Cards help train the model.


How suggestions improve over time

When AI Suggest is first enabled, Guru uses basic content similarity to recommend Cards. Over time, the model adapts to your team’s behavior.

Actions that help improve suggestions include:

  • Copy actions:
    • Copying text, code blocks, or images from Cards.
    • Using Copy Card URL or Copy content.
  • View actions:
    • Opening a Card.
    • Switching tabs in the Card pop-out window.

Guru processes these actions multiple times per week to update the model.


Setting boundaries for suggestions

Enabling AI Suggest means Guru may process limited data from enabled apps. All data is stored securely and only used to power suggestions.

Group limiting

  • Restrict suggestions to specific user Groups.
  • If left empty, all Groups will receive suggestions.

Collection limiting

  • Restrict suggestions to specific Collections.
  • If left empty, suggestions come from any Collection the user has access to.
  • All permission rules still apply.

Data masking

  • Prevent storage of sensitive information by applying masking rules.
  • By default, Guru masks email addresses, phone numbers, and social security numbers.
  • Admins can add custom regex masking rules.**

Viewing suggestion analytics

Admins can track usage by visiting the Adoption and Feature Usage tab in Analytics.
Learn more in Evaluating Adoption and Feature Usage Analytics.


Frequently asked questions about AI Suggest

Do I need the browser extension to use AI Suggest?
Yes. The extension collects data and displays up to 5 suggested Cards.

Can I limit what Cards are suggested?
Yes. Admins can limit suggestions by:

  • Specifying which Collections to search.
  • Restricting access by Group.

How quickly will suggestions appear after enabling AI Suggest?
It may take up to 24 hours to generate your first suggestions while the ML model is initialized.

How soon do new Cards appear in suggestions?
New Cards appear within 3–7 hours of being published.
For immediate alerts, consider using Knowledge Triggers.

How many apps can I enable AI Suggest for?
You can enable AI Suggest for any number of supported apps from the AI Suggested Text settings page.

How can I improve suggestions faster?
Encourage team members to:

  • Use the browser extension frequently.
  • Revisit past conversations in enabled apps and interact with relevant Cards.

Can I control what information Guru reads?
Yes. AI Suggest only processes fields necessary for suggestions. Admins can:

  • Use built-in masking for emails, numbers, etc.
  • Add custom masking rules for more control.

Does each app have its own ML model?
No. Guru maintains one model per customer, regardless of the number of apps enabled. The model adapts based on all user interactions across tools.

Does AI Suggest work in other languages?
AI Suggest performs best when both conversations and Cards are in English. Quality may decrease with other languages.