Our Agentic Platform helps your team move beyond basic chat prompts to automate repeatable research workflows using governed agents. From the Agents page, you can create new agents, configure their instructions, and connect them to the platform and external data. This article explains how to create an agent, update its configuration, manage versions, and reuse or share agents to consistently generate rich, review-ready deliverables like slide decks, spreadsheets, comparison tables, and reports.
This article will cover the following topics:
If you are looking for information regarding the Research Assistant chat feature, please see the dedicated article below:
Agents are governed by two permissions: Agent Create and Agent Manage All. These permissions follow from their names: Create grants the ability to add agents while Manage All grants the ability to create and modify any agent. A System Administrator will need to grant these permissions before you can start creating agents.
Agents can be found within the Research Assistant module. You will see a new toggle for Chat and Agent when you open Research Assistant. Following the steps below will allow you to see a table view of all available agents.
-
Navigate to the Research Assistant module.
-
Switch to Agent Mode.
-
Select View Agents.
Users with the appropriate permissions can also create new agents or manage existing agents from this view.
You can manage agents directly from the Agents table or Details page.
From the Agents table, use the 3-dot menu to update basic properties such as the name or description for a given Agent. After saving, the table will refresh with the updated values.
Agents can be shared with other members of your organization to allow them to execute the agent as needed. You can share with all users or individual users. Shared agents can also be revoked from users from this option.
Clone Agent can be used when you want to reuse an existing setup. You can only duplicate a published version.
Each agent can have multiple versions, allowing you to track changes over time.
To view agent versions:
-
Open the Agent Details page.
-
Open the Actions menu and select Versions.
The Versions page shows version details such as:
-
Version Number
-
Status
-
Created On
-
Modified On
-
Created By
After clicking the Version Number to view the details page, you can take several actions including:
-
Restore: On inactive versions you can restore the version to published. This will set the current published version to inactive.
-
Clone as Draft: On inactive or published versions you can clone the version to a new draft version.
-
Save & Publish: You can save draft agents that are in the process of being configured to continue their setup at any time. You can publish any draft version. If you are cloning as a draft, publishing will set the current published version to inactive.
You can either create an agent from scratch or from your library of agents. Creating from the library saves you time, allowing you to select from a list of curated agents created by your SME team.
-
From the table view select Add Agent.
-
In the modal, enter the Name and Description.
-
Click Save to create the Agent.
-
The page will reload and show the details screen for your new agent.
-
From the table view select the dropdown next to Add Agent.
-
Click Clone from Library.
-
Select the Agent you wish to clone.
-
Click the Select and Close option.
-
The table will refresh with the cloned agent. You can click the name to open details and customize this agent.
Note: Standard users can clone and run agents shared with them. Privileged users (such as admins or SMEs) have the ability to publish custom agents to the library for the wider team to clone and use.
In the Agent Details page you can modify what data agents have access to and update their instructions. You can also publish changes and create new drafts.
-
The Connectors tab lists the available data sources an agent can read. External data sources must be first configured in the System Admin Module: Integrations tab.
-
The Instructions tab shows the instructions the agent follows on run.
-
The header row shows the status of the current agent version. Only draft versions can be edited.
-
To edit a published version, click Clone As Draft to make a new draft version then publish that version.
-
When you publish a version it will become the active version and the prior active version will become inactive.
-
Note: If a user has shared an agent with you but you do not have Manage All, you will not be able to edit any of these settings.
Data connectors control what data an agent can access during execution. Some connectors may appear locked and disabled. These connectors must be first configured in Integrations under System Admin. Connectors are categorized as either user-level or company-level integrations.
-
The web connector shows an option to limit the connector to specific websites.
-
The Outlook and Google Drive connectors are user-level integrations and will prompt the individual user to authenticate their personal Outlook or Google Drive account when running the agent. Please see our articles on these integrations below:
-
SharePoint and Salesforce are company-level integrations that utilize a single integration user profile with specific read/write directory permissions configured by an administrator. Please see our articles on these integrations below:
-
The Adhoc File Picker will prompt users to add files when running the agent.
You can find more information regarding the integrated AI agent connectors in the article below:
Instructions tell an agent the precise workflow to follow including what entities are in scope (Partner, Products, Funds, etc.), what steps to follow, and what outputs to deliver. Instructions must be written using Markdown syntax (such as hashes for headers or asterisks for bold text) so the agent can properly interpret structure and logic steps.
Instructions should describe:
-
What user inputs are required (i.e., does the user need to provide a list of Partners or Products?).
-
The goal of the agent.
-
The sequence of steps it should follow.
-
Any decision logic it should use.
-
The type of output it should produce.
An Optimize button is available next to the instructions box. Clicking this will automatically apply prompt engineering best practices to refine your instructions, standardize tone, inject safety boundaries (e.g., ignoring unverified sources), and detail exactly what was changed.
You can save agents that are in the process of being configured to continue their setup at any time by clicking the Save button. Once configuration is completed, you need to publish the agent to begin using it by clicking the Publish button.
After you publishing an agent, you can start using it. More information on running an AI agent can be found in the article below: