What Enterprise Search is and how its admin console is organized. Each section below answers one question and stands alone.
What Is Enterprise Search?
Enterprise Search lets people ask work questions in plain language and get AI answers drawn from your organization's connected, access-controlled knowledge — with citations back to the source. It has two surfaces:
The search experience (Enterprise Search in the workspace) — where employees actually search and get answers.
The admin console (Enterprise Search admin) — where you configure how search behaves: its instructions, models, connected apps, governance, the people it knows about, and its analytics.
Both are powered by one behind-the-scenes Enterprise Search agent, so configuring the admin console changes how the search experience answers.
What Is the Enterprise Search Agent?
Under the hood, Enterprise Search is a single AI agent dedicated to search. That's why the admin console looks so much like the AI Agent editor — it's configuring a real agent, just one that's wired to the search experience. You don't create or pick this agent per page; the console always configures the one agent behind Enterprise Search for your workspace. If it hasn't been set up yet, the console shows an empty state until the agent exists.
How Is the Admin Console Organized?
A left side-nav groups every setting, top to bottom:
Section | Covers |
|---|---|
Instructions | The agent's role, goal, and behavior, worded for search. |
Knowledge | The sources the agent searches: applications, documents, websites, and data catalog. |
Connectors | The apps Enterprise Search can act through, and what it's allowed to do in each. |
Models | The AI models behind each stage of search. |
Memory, Skills, Capabilities | What the agent remembers, the expertise attached to it, and the built-in functions it can use. |
Insights | Usage, Feedback (Summary / Analysis), and Spending (By User / By Model). |
Observability | Sessions, Traces, Spending, Memories, Evaluation. |
People | The Directory, its Sources, and its Schema. |
Governance | Incidents, Policies, and Controls, each split into Chat and Knowledge. |
Settings | Indexing, Pre-processing, Response generation, and Prerequisite Actions. |
Test agent (under Agent preview) | A test chat inside the console, for trying the agent with your unpublished changes. |
A persistent header across the top shows the title Enterprise Search with the subtitle "Admin · insights, people, and governance for Enterprise Search", plus Preview app and Publish. Preview app opens the live Enterprise Search chat app in a new tab.
Which Sections Are Unique to Enterprise Search, and Which Mirror the AI Agent Editor?
Most config sections are the same screens as the AI Agent editor, just mounted for the search agent — Models, Memory, Skills, Capabilities, Observability, People, and everything under Settings (Indexing, Pre-processing, Response generation, Prerequisite Actions). Instructions is the same screen with search-specific wording.
The sections unique to Enterprise Search are Knowledge, Connectors, Governance, Insights, and the end-user search experience.
How Do I Publish Changes, and What Does "Deploy on Copilot" Do?
Some sections (such as Memory) save automatically. Others (such as Instructions) have their own Save button. Either way, your changes only reach the live search experience when you Publish. Publishing needs deploy permission and opens the standard publish dialog, which includes a Deploy this agent on Copilot checkbox (on by default; after the first publish it reads Deploy the new version on Copilot). Leave it on so the new version becomes the one the Enterprise Search experience actually uses. Publish once your instructions, connectors, and settings are ready.