Modes, Feedback, and Tips

How you set up and phrase a request shapes how well Autopilot does it. This page covers the two reasoning modes and the thinking effort setting, the two ways to view Autopilot's work, how to attach files, how to give feedback, the keyboard shortcuts, and instructions you can copy to get started.

Fast and Thorough modes

Autopilot offers two reasoning modes. You choose one per conversation, before you send your first message, using the Thinking mode control (the pill in the bottom-left of the composer, next to the plus (+) button).

• Thorough — the default. Autopilot reasons more deeply and surfaces its thinking and tool steps. Selecting the pill opens a thinking effort slider: Off, Minimal, Low, Medium (the default), or High — that sets how much room Autopilot has to reason before it acts. Higher levels are slower but better on hard, multi-step problems; stay at Medium or above when you are building, editing, or refining an AI Employee.
• Fast — quicker responses with streamlined reasoning. Turn on the Fast mode switch in the same control. While Fast is on, the effort slider is hidden. Fast uses a quicker model that doesn't apply thinking effort. Use it for simple, single-step requests like looking up information or listing AI Employees.

The collapsed pill shows your current setting: in Thorough mode, the thinking effort level with a small bar meter (for example Medium), or Thinking off; in Fast mode, Fast with a small lightning bolt. Your choice is remembered across sessions.

Narrative and Developer view

The view setting controls how much detail Autopilot shows about its work. Change it from the More actions menu (the three-dots button in the panel header) under View mode.

• Narrative — the default. Autopilot's steps are grouped into clear, human-readable summaries ("Searched knowledge base," "Created workflow," "Completed 3 sub-tasks") so you can follow along without technical detail.
• Developer — shows the individual tool steps verbatim, each with its inputs and outputs, plus Autopilot's step-by-step reasoning. Use it to debug an AI Employee's behavior or to see exactly how Autopilot carried out a task.

Both views show reasoning (Autopilot's thinking, when the mode produces it) and tool steps (the actions it takes); Developer view simply shows them in full rather than summarized. A long tool output is previewed with a Load full output control to expand it.

Your view choice is remembered across sessions. (The thinking effort setting lives in the composer's Thinking mode control — see Fast and Thorough modes above.)

Uploading files

Attach a file to give Autopilot extra context: a policy document for an AI Employee to learn from, or a spreadsheet of test questions. There are three ways to attach:

1. Select the plus (+) button in the composer to open your file picker.
2. Drag and drop files into the panel. A highlighted border appears to show the drop zone is active.
3. Paste an image directly into the composer.

Attached files appear as chips above the composer showing the filename and size, with an X to remove one before you send. After attaching, tell Autopilot what to do with the file, for example, which AI Employee the data belongs to.

Giving feedback

Your feedback helps improve Autopilot. Hover over any Autopilot reply to reveal an action row:

• Thumbs up — marks the reply as helpful and saves it right away.
• Thumbs down — opens a short form where you can pick a reason and add an optional comment, then submit. The reasons are: Wrong or lacking information, Not what I requested, Laggy or malfunctioning, Safety or legal issue, and Miscellaneous.
• Copy — copies the reply text to your clipboard.

Feedback is per reply, so you can rate individual responses within the same conversation. Selecting a thumb again clears your rating.

Keyboard shortcuts

You can drive Autopilot without leaving the keyboard.

Workspace-level Autopilot configuration

Workspace admins can set a default configuration that applies to every unbound Autopilot session in the workspace. Open Workspace Management and select the Autopilot tab. Changes take effect on the next turn of any active conversation. No restart is needed.

Three settings are available:

• Additional instructions — free-text instructions appended to Autopilot's system prompt for every default session. Use this to add workspace-specific guidance, terminology, or behavioral rules that apply across all users. These instructions do not replace Autopilot's built-in identity; they are appended after it.
• Skills — workspace-global skills visible to every default Autopilot session in the workspace. Skills added here are not visible to bound orchestrator AI Employees.
• Web access — enables or disables Autopilot's ability to search the web. When turned off, both web search and related tools are removed from default sessions workspace-wide.

Example instructions

When you open a new conversation, the welcome screen shows suggestion cards that pre-fill the composer when you select one. The starting set covers the most common first tasks:

Beyond the cards, here are instructions you can paste and adapt:

List the most recent 5 workflows in my workspace.

Search my knowledge base for our refund policy and summarize the top results.

Draft a one-page onboarding guide covering our PTO policy, expense process, and IT setup checklist. Use headers and bullet points.

For the complete set of actions you can ask for, see Core actions.

What's next

• Core actions — everything Autopilot can build and change.
• Permissions and scoping — what Autopilot can access and when it asks for approval.
• Limits, availability, and troubleshooting — what Autopilot can't do, and what to do when it's unavailable.