Copilot Agents – Example Agents

Page contents

• Introduction
• How to add the example Agents
• Examples – Copilot Agents that answer questions based on custom knowledge
• Examples – Copilot Agents that create or format text in a specific way
• Examples – Copilot Agents as learning activities

Introduction

Return to page contents

On this page there are links to some example Copilot Agents that the Ed Tech team have created to demonstrate the potential of these ‘custom chatbots’. The Agents are divided into three categories to showcase different ways Agents can be used.

Each example Agent also has the ‘prompt’ (instructions) used to create it, which you can download as a Word doc. You may find it helpful to look at these prompts as a reference when creating your own Agents.

See Copilot Agents – How to create and share for further guidance for creating your own Agents.

How to add the example Agents

Return to page contents

To use one of these Agents, click the links in the grey drop-down boxes below, sign in to your Coventry account if you aren’t already, then select ‘Add’ on the pop-up, then ‘Open’. The Agent will then be available in the side bar of the Copilot web app (you can pin it with the 📌 symbol if you want to keep it at the top).

Tip: Some people have reported errors when trying to add Copilot Agents. If you have selected ‘Add’, but then when you select ‘Open’ nothing happens, try going directly to ⁠Copilot to see if the agent has been added (it will appear in the side bar). For support or to report broken links, please email the Ed Tech team at learningehance@coventry.ac.uk

Copilot Agents that answer questions based on custom knowledge

Return to page contents

The instructions for your Copilot Agent can be up to 8,000 characters, and within this you can include ‘knowledge’ that you want it to base its answers on. This can be knowledge that a normal AI wouldn’t have, e.g. information about your course or team. See Copilot Agents – How to create and share for guidance about how to write instructions for Copilot Agents.

Here are some examples of Copilot Agents that answer questions based on custom knowledge, along with the prompts (instructions) used to create them:

Copilot Agents that create or format text in a specific way

Return to page contents

One good way to use Copilot Agents is to create Agents that can speed up repetitive tasks that you have to do regularly (such as creating meeting notes) or standardise common tasks that are done by several members of your team (such as creating quizzes).

It’s important to know that the version of Copilot we have is not good at working with long documents or big data sets. It only works well with a page or so of text. See Tips for using Copilot to learn more about how it compares to other AI tools.

Here are some examples of Copilot Agents that create or format text in a specific way, along with the prompts (instructions) used to create them:

Copilot Agents as learning activities

Return to page contents

Without guidance, students may not use AI in a way that promotes learning. To help raise their awareness of how to use AI positively, you could create Copilot Agents for lots of different types of learning activities: an Agent that leads a role-play, an Agent that explains concepts in a way that’s suited to the individual user, an Agent that provides concrete worked examples, an Agent that quizzes students on one topic and provides timely feedback, an Agent that quizzes students on all the topics covered so far using interleaving and spaced practice principles, an Agent that asks students to teach it about what they’ve been learning, an Agent that leads a guided reflection … You could make a different ‘learning Agent’ for each week of your course!

If you provide ‘knowledge’ in the instructions for your Agent that it can use in it’s questions and feedback for the learner, you can make sure its answers are appropriate-level and accurate, which is not always going to be true with more general AI tools.

There are a few important things to be aware of when creating learning activities. In your instructions, I recommend that you:

• Explicitly tell Copilot whether it can ONLY use the information you provide, OR if it can also use its general knowledge.
• Get Copilot to explain the limits of its knowledge to the user before the activity begins, so they know not to ask it questions beyond its scope.
• Tell Copilot the format for its suggested answers (it’s not possible to turn these off, and unless you tell it what to put in them, it will just give away the correct answers every time). E.g. If you make a Copilot Agent that quizzes students with multiple-choice questions, you could tell it: ‘If you suggest responses for the user, always suggest: A, B, C. This ensures you aren’t inadvertently revealing the answers to the quiz.’

The ‘Ed Tech team quiz’ example below demonstrates these tips in practice. You can have a go at customizing this prompt to make your own Agent with this interactive web page: 🔗 ‘Team Quiz’ – Prompt edit activity.

Here are some examples of Copilot Agents as learning activities, along with the prompts (instructions) used to create them: