Distributed cognition is a well-established concept that suggests thinking does not just happen in your head. It extends across people, tools, and the environment around you (Hollan, Hutchins, & Kirsh, 2000). Your brain works together with things like your phone, notes, other people, and everyday objects to solve problems and make decisions.
We have always used this principle across everyday life and professional settings, including classrooms, offices, and workplaces, even if we did not know it was called distributed cognition. The goal has often been the same: offload effort where possible so we can focus our time and energy on more meaningful tasks.
A simple example is using a calculator. We still learn basic math because understanding the fundamentals matters. But when it comes to complex calculations, we rely on calculators to save time and reduce cognitive load. The tool does not replace our thinking. It supports it. In a similar way, Copilot Agents can provide a more work-specific extension of that same idea.
Copilot Agents are tailored AI tools designed to support specific tasks. Instead of being only a general-purpose AI assistant, an agent can be built with instructions, context, and knowledge related to a specific kind of work. You can think of them as structured support tools for repeated information tasks, especially when the goal, sources, and boundaries are clear.
Thinking of Copilot Agents as cognitive partners
If you regularly find yourself repeating the same explanations, answering similar questions, or guiding people through standard processes, that is a strong signal that a Copilot Agent could help. These repeated patterns indicate that some of the thinking involved, including recalling information, interpreting requests, and forming responses, is consistent enough to be supported by a tool.
Imagine an agent that provides the same type of guidance you would, designed to be consistent, aligned with your instructions, and available when needed. It can remove some of the repetitive layers of your work while still preserving your expertise.
A good analogy is navigation. Most of us no longer rely only on memorizing routes or using a road atlas. We use GPS. It is still useful to know common routes, but GPS saves time and effort, especially in unfamiliar situations. It allows us to focus on driving rather than constantly figuring out directions. In that sense, GPS becomes part of the cognitive system. It takes on tasks like route planning and direction support, which would otherwise rely more heavily on memory and reasoning.
Copilot Agents work in a similar way. They do not replace your knowledge. They can make repeated knowledge work more consistent, scalable, and accessible.
A real example from ITS: HawkAI Level 1 Agent
Within the AI Support Team at ITS, we faced a recurring challenge with the HawkAI Level 1 course series. We were receiving many similar questions from users, often requiring us to provide the same guidance each time. Over time, a significant portion of HawkAI support involved answering these questions and guiding users through the same process.
That pattern made it clear that the underlying information and responses were consistent and well-suited for support through a Copilot Agent.
To address this, we built a Copilot Agent. We configured the agent with:
- Session descriptions
- Course schedules and offerings
- Instructor details
- Certificate and quiz requirements
- Internal guidance rules we typically used when responding
We also refined the agent over several iterations so its responses were clear, consistent, and aligned with our guidance.
Once it was shared with HawkAI Level 1 participants, the agent gave users another place to start with common questions before contacting the team. It helped move some routine information lookups into a consistent, self-service format.
This allowed the team to focus on program improvements and complex support needs.
Where else can Copilot Agents help?
The same idea applies to many other areas of work. Copilot Agents can support tasks such as:
- Drafting emails and communications
- Refining reports or other documents with feedback
- Analyzing data and summarizing insights
- Generating action items from meeting notes
- Answering common questions using approved public websites or shared guidance
- Providing structured guidance on internal processes
Any task that is repetitive, time-consuming, or cognitively demanding may be a strong candidate for an agent, especially when the information source is clear and the output still receives human review.
Why this works
The value of Copilot Agents comes from a simple principle: offloading.
When you delegate certain tasks to a system designed to handle them consistently, you free up time and mental energy. That cognitive space can then be used for higher-value work such as decision-making, problem-solving, and strategy (Sweller, 1988; Hollan, Hutchins, & Kirsh, 2000).
Think again about GPS. You are not spending as much time interpreting maps or second-guessing directions. Instead, you can focus on driving safely and reaching your destination efficiently. By taking over tasks like route planning and navigation support, GPS reduces the need to rely only on working memory or previously learned routes.
Copilot Agents can do something similar for knowledge work. They can help with the repetitive and structured parts, allowing you to focus more attention on the parts that require human judgment.
Responsible use
Copilot Agents can help draft, organize, and summarize information, but their output still needs to be reviewed. Be careful about the data, documents, or websites you connect to an agent, and follow university guidance for the type of information you are using.
Agents should be treated as support tools, not final decision-makers. They can help retrieve information, structure responses, and reduce repetition, but people are still responsible for checking accuracy, context, data sensitivity, and policy alignment before using or sharing the output.
Getting started with Copilot Agents
At Iowa, Copilot Chat Agents are available to UI students, faculty, and staff through Copilot Chat. You can learn more about building and using them on the Copilot Chat Agents support page.
Faculty and staff with a Microsoft 365 Copilot license may have additional options, including the ability to ground agents in Microsoft 365 files and folders they already have permission to access. Availability and features can vary, so check the AI Tools page and current ITS guidance before building an agent for shared use.
To explore how Copilot Agents can be useful in your own work, you can register for a live training session.
This session is available for everyone at the University of Iowa.
Final thought
Think of Copilot Agents not just as tools, but as cognitive partners. The examples discussed here highlight only a few ways they can support your thinking. There are many other ways to use them to extend and augment how you work.
When used thoughtfully, agents can become part of how you think, work, and collaborate by supporting tasks like retrieving information, structuring responses, and guiding people through repeated processes. The goal is not to replace human effort, but to redistribute it in a way that makes work more effective.
The question is not whether you should use agents everywhere. The better question is where in your work they can make the biggest difference.
To learn more about AI tools and guidance at Iowa, visit the AI Tools page and the Iowa AI Hub. You can also subscribe to the AI at Iowa newsletter for updates, events, and training opportunities.