Copilot Agents in SharePoint
Like what you see??
"Ask Sympraxis" is a bi-weekly webinar series, where we discuss an array of topics and answer your submitted questions. Join us by downloading our recurring calendar event. You can also join us directly in the meeting without downloading the event.
See a listing of Ask Sympraxis episodes by topic covered: Topic List, Series List, or a full listing Archive
In the rapidly evolving landscape of AI-powered tools, SharePoint Copilot agents stand out as a game-changer for enhancing workplace productivity. These specialized AI assistants integrate your organization’s knowledge and data sources with Microsoft Copilot, enabling users to perform a range of tasks more efficiently. From automating repetitive processes to offering actionable insights, Copilot agents can transform how teams interact with data and make decisions.
What Is an Agent?
A Copilot agent is an AI assistant tailored to connect your organization’s data sources to Microsoft Copilot. By integrating with your data, agents empower users to access the right information at the right time, enhancing productivity across the board.
How to Create an Agent
Creating an agent begins with choosing the right location. Just as you’d search for information at its source, your agent should be created close to the data it will use. For example, if your agent will assist with HR-related tasks, set it up near your HR data. Agents can be created at the site level, where they’re stored in the site assets folder by default and scoped to that specific location.
If you create an agent within a hub site, it will automatically target all the sites connected to that hub. The takeaway here: ensure the agent’s scope aligns with the data and users it’s intended to serve.
Getting Started
Setting up an agent is straightforward. Use the “Create an Agent” option in the menu bar or click the Copilot button in the suite bar. This launches a dialog box that guides you through the process. Once created, your agent is ready to use with just one click. Customizing your agent is equally simple. You can: rename the agent for clarity (e.g., change “Documents Agent” to “Helen from HR”), add a description to help users understand its purpose, adjust the data sources the agent uses to respond to queries, and define the agent’s behavior and interaction style through the behavior tab.
Advanced Customization in Copilot Studio
While Copilot Studio holds promise for advanced customization, many features are still marked as “coming soon.” Future updates may include importing solutions beyond individual files, expanding the potential for agent capabilities.
Things to Consider
Effective agents rely on high-quality content. If your organization’s information architecture or content organization is flawed, Copilot agents will only amplify these issues. Ensure your documents are well-organized and provide clear answers to anticipated user prompts. Additionally, take care when naming agents, as their names will appear prominently in the user interface.
It’s also worth noting that users with access to Copilot might see agents across site collections, depending on permissions. Properly scoped and named agents can help streamline content discovery and interaction.
Approved Agents
Any member with edit access to a site can create and use an agent, but only site owners can approve them. Approval makes an agent available across the site and allows it to be set as the default agent. Before approval, the agent file resides in the creator’s site assets folder. Once approved, it moves to the Copilots Approved folder in site assets.
Copilot Agent Scenarios
Potential use cases for Copilot agents are vast. For example, they could enhance interactions with policy libraries by allowing conversational access to policy documents. While Microsoft has shared a variety of scenarios, the success of these agents hinges on the quality and relevance of the content they access. Without well-vetted information, agents may fall short of delivering on their potential.
Provisioning Agents
Agents are stored as JSON files, with the filename reflecting the agent’s name. Within the JSON file, the “capabilities” section specifies the lists and libraries the agent interacts with. This simple structure allows for flexibility and customization.
By integrating seamlessly with your organization’s data and providing intuitive user interactions, they have the potential to redefine how teams work. However, their effectiveness depends heavily on the quality of your content and data organization. As this technology evolves, investing in good information architecture and thoughtful agent design will be critical to unlocking its full potential.
All Resources
Do you have any questions for us? Continue the conversation on BlueSky with the hashtag #AskSympraxis and mention @sympraxisconsulting.com.