Skip to main content

​
Overview

With in-line execution enabled, eligible write actions in interactive agents can execute automatically without showing a confirmation step. This removes extra clicks and shortens the time to complete tasks, especially in workflows where confirmation is not necessary. When configured by an admin and enabled by the agent creator, the agent executes the write action as part of the normal conversation, reducing friction while preserving safety controls defined for the action. In-line execution respects the safety rules and action constraints of your organization:
  • Admins determine which actions are allowed to run without confirmation.
  • Agent creators must explicitly opt in at the step level.
  • Standard permissions, authentication, and audit logs continue to apply.
  • If an action or step is not eligible, the agent either prompts for confirmation (if allowed) or surfaces an error.
After you enable in-line execution of actions:
  • No confirmation step: In eligible flows, the agent runs the write action without asking the user to click Confirm.
  • Single, streamlined flow: The agent proceeds to the next step or returns results as part of the same conversation.
  • Normal responses on completion: The agent summarizes the change or next steps like any other action result. Errors are surfaced with guidance to fix inputs or permissions.
When actions are enabled to run without user confirmation, Glean automatically executes the action based on AI-predicted values and might update the system of record. We recommend that agent builders thoroughly test action outputs and guardrails in a controlled environment before broad rollout.

​
Configuration steps

To use the in-line execution of actions feature, you must first configure this feature in the Admin console and then enable it in the Agent Builder. Some creation or drafting of actions may be enabled to run without confirmation by default, while others require an admin to opt in. As defaults can change over time, check the settings of specific action in Admin console.

​
Configuring in Glean Admin console

​
Prerequisites

  • Admin access to Glean.
  • The action must be a supported write action, for example, updating a record in a third‑party system.

​
Steps

Perform the following steps to enable in-line execution of actions:
  1. Navigate to Glean Admin console.
  2. Click Platform → Actions.
  3. Click the action pack you want to configure and go to Configuration tab.
  4. Click Edit settings under Enable actions.
  5. Click Agents and set Run without user confirmation to Yes to enable in-line execution of actions.
  6. Click Save.

​
Enabling in the Agent Builder

Agent creators can decide whether to run a configured write action in-line within a specific agent. Perform the following steps to enable in-line execution of actions configured in the Admin console:
  1. Open the agent in Agent Builder.
  2. Add or select the step that runs the relevant write action you configured in Admin console.
  3. Enable the option to run the action in-line.
  4. Save and publish your agent.
End users must grant consent for an agent to execute allowed actions without a confirmation step. During consent, users can review what permissions or actions the agent may perform. Glean stores this consent for future runs until the agent version is updated, at which point users will be prompted to re‑consent.

​
Known limitations

  • Not supported for custom actions.
  • Applies to supported write actions only.
  • Designed for interactive agents (chat-based flows).
  • Some actions may always require confirmation due to risk or policy.
  • Tenant policies, authentication state, or missing configuration can block in-line execution.

​
Troubleshooting

If a write action still shows a confirmation step or fails to run in-line:
  • Confirm the admin has enabled the action to run without confirmation.
  • Verify the agent step is configured to run in-line in Agent Builder.
  • Check user permissions and authentication for the destination system.
  • Re‑test with a simpler change or a single record to isolate issues.
  • Review agent step instructions for clarity and guardrails.

​
FAQs

No. In-line execution follows the same permission checks, admin controls, and auditing as standard actions.
In-line execution must be explicitly allowed by admins at the action level and then opted into per agent step by the agent creator.
The agent returns an error with guidance to fix inputs, configuration, or permissions, and you can re-run the flow after addressing the issue.
Yes. Actions executed in-line are recorded like any other action run in your environment.
⌘I