How to integrate generative AI into your courses

Classroom AI Ideas from Easy to Advanced

The examples on this page are grouped by the level of time and difficulty. Don't let that intimidate you, though. Even the most difficult approaches are achievable.
Easy AI Course Integration Easy AI Course Integration
Intermediate AI Integration Intermediate AI Integration
Advanced AI Integration Advanced AI Integration

A growing number of instructors have found innovative ways to integrate generative artificial intelligence into teaching and learning. Their experimentation provides ideas and models that other instructors can draw on for turning generative AI into a tool for learning in their own courses.

Adding assignments related to generative AI doesn’t have to be complicated, and many of the examples below can be copied and easily adapted for your own class. The approaches have some common themes:

  • Student experimentation. Instructors have created guided exploration of generative AI tools as a way to help students build skills in prompting and critical evaluation, and to help them better understand how to use generative AI for learning.
  • Critical evaluation. AI-related assignments that include reflection, writing, or some form of evaluation of chatbot output help students learn to approach generative AI skeptically, building critical thinking skills.
  • Connection to specific materials. An approach known as retrieval-augmented generation, or RAG, provides specific materials for generative AI systems to draw on. That approach improves the accuracy of responses and helps create a type of custom chatbot. 
  • Custom instructions. These are essentially prompts embedded in custom chatbots (like Copilot agents). They guide chatbots by telling them how to act (and how not to act), what types of materials to draw on, and what types of output to provide. 

Below are some examples of how other faculty members have used generative AI in their classes. Most emphasize use of Microsoft Copilot, which is the most secure environment for KU faculty and students. Other tools have many possibilities, though, and can be used with many types of assignments.

Don't be afraid to experiment. Find ways to help students learn to use generative AI critically, and talk frequently with students about when to use it and when not to use it for learning.

Resources on using Copilot and Copilot agents

Easy AI assignment approaches that require little time

Students can critically evaluate AI-generated content by fact-checking informatio n, identifying biases, and comparing outputs. This less time intensive approach helps them learn course material, strengthen critical analysis skills, and recognize the weaknesses of AI-generated content.

Critiquing AI Responses and Outputs

Tara Marriage in undergraduate biology at KU created an assignment related to hydrogen bonding, a concept students often struggle with in introductory biology courses. Here’s what the assignment involved:
Part 1: Create an AI-generated explanation

  • Have Microsoft Copilot explain a specific aspect of hydrogen bonding you find challenging.
  • Analyze those explanations for inaccuracies and patterns.

Part 2: Create analogies

  • Have Copilot to create five analogies explaining hydrogen bonding from different perspectives.
  • Evaluate each analogy for scientific accuracy.

Part 3: Create illustrations

  • Use Copilot to create three visual representations of hydrogen bonding.
  • Choose the most effective one and reflect on how on its accuracy and on the components of the illustration.
Critique of AI-generated writing 

Students prompt generative AI to write an essay on a topic related to some aspect of a class. Then they critically evaluate the accuracy, biases, perspectives, and logic. This approach can help students learn course material and disciplinary thinking. It also helps them look beyond the surface of AI-generated content and recognize its weaknesses.

Fact-check AI output

Students prompt generative AI to write an essay on a topic related to some aspect of a class. Then they critically evaluate the accuracy, biases, perspectives, and logic (including material that sounds plausible is inaccurate or illogical). This approach can help students learn course material and disciplinary thinking, as well as strengthen their critical analysis skills. 

Prompt: Act as an expert on students, faculty and staff at the University of Kansas. Draw on web and social media sources to create a profile of (name and department). Where have they studied? What are their specialty areas? And what can you tell me about them as teachers and as scholars?

Fact-check AI-generated citations

Students use a generative AI tool to create a bibliography on a course-related topic. They then verify the sources with library databases, Google Scholar, and web searches, identifying inaccurate or fabricated publications.

Identify biases in AI-generated images

Have students use Bias Lens, a generative AI tool created by CTE, to explore the biases inherent in AI-generated images. The tool asks students to enter a generic title like “doctor” or “engineer” or “teacher.” It generates an image and asks students to reflect on characteristics and appearance before providing guidance on writing more effective prompts. Students are then asked to generate another image, compare that to the original, and reflect on the process.

Compare output from generative AI tools

Students create a prompt that they use in two or more generative AI tools. They then compare the output of each, analyzing them for differences and discussing which is best and why.

Review and debug AI-generated code

Students use a generative AI tool to create code snippets. They then evaluate the output for inconsistencies and errors, identifying ways to improve the prompt they used. After rewriting the prompt, they create the code again and compare the result with the original.

Where to find other examples

You will find other assignments on the sites of the AI Pedagogy Project,

Create a basic chatbot in Microsoft Copilot

Doug Ward in journalism and mass communications assigned students to create their own Copilot agent, which is a personalized version of Copilot. Students were told that the agent should focus on some aspect of their journalistic work where they need feedback or assistance.

To help with the process, he created an Agent Idea Generator with Copilot. That tool helped students brainstorm ideas, provided examples of Copilot agents, and helped them refine their ideas. (The idea generator will work for any discipline.)

Here were the instructions he gave to students:

Part 1: Ideation

Part 2: Create an agent

  • Once you have settled on an idea, go to Copilot and log in with your KU credentials.
  • Click on Create Agent.
  • Explore the templates available for the agents. What can you learn from the structure of the instructions?
  • Describe the type of bot you would like to create using the chat function at the bottom of the page.
  • Create your bot and experiment with it, refining the instructions as needed. The Copilot Prompt Coach can help.   

Part 3: Reflect

Submit a document that includes the name of your chatbot, a description of what it does, and the instructions you gave it. Then provide a paragraph of reflection: 

  • Did anything surprise you about the process?
  • How does this differ from the way you have used chatbots in the past?
  • Does it give you ideas for how you might create another tool in the future?
Integrate CTE modules on AI literacy into courses

These modules, created in Canvas, provide a basic understanding of what AI is, how it can be used, why it must be viewed critically, and how it is affecting society. They encourage ethical use of generative AI, helping students understand its many shortcomings and biases (and strengths and benefits) so they can make reasoned judgments about when or whether to use it. Contact CTE if you are interested in getting access to the modules. 

Have students create research maps

Several online tools allow researchers to use generative AI to create maps of interconnected research and to identify related research. Each works in a slightly different way, but all are intended to help researchers make connections among ideas, find additional sources, and consider new ways of thinking about a topic. 

Assignment: Based on your research interests, create a research map or a report on related research using one of these tools: Research Rabbit, Scite, Semantic Scholar, Perplexity, Elicit, or Connected Papers. In class, be prepared to share your visualization or report and talk about how you used the tool, whether you found it useful, and whether it helped you find resources you might not otherwise have found. Also be prepared to discuss the accuracy of the results. 

Intermediate AI Integration

Chatbots can be customized for various personas and scenarios, making them useful for class activities. Examples include mock patient interviews, pharmacy coaching, occupational therapy simulations, and startup pitch practice.

Critiquing AI Responses and Outputs

NotebookLM is a free Google tool that allows you to ask questions of a collection of sources you give it. It can also create overviews, podcasts, videos, and visualizations of those sources. Doug Ward in journalism and mass communications created a notebook with materials on how journalists are using generative AI. He had students watch a video and listen to a podcast NotebookLM created from those sources. He then asked them to reflect on the materials and engage in a class discussion: 

  • How is NotebookLM condensing ideas from all the sources?
  • What is it leaving out?
  • Can you envision journalistic uses for a tool like this?

Compare the approach of the podcast with that of the video:

  • What are the differences?
  • Why are they different?
  • What did you learn from the podcast that you didn't from the video?
  • Which aspects of the video or audio would you like to learn more about?

This assignment takes a bit of time to set up NotebookLM. Once you have the notebook, though, it can be a great resource for a class.

Doug Ward in journalism and mass communications created an assignment in which students used Google Gemini to create an interactive app.

Part 1: Explore an example app

Go to Bias Lens, an interactive app I created with Gemini. Work through the image generation and the reflection prompts. The app is an example of the type of interaction you can create without coding anything yourself. 

Part 2: Brainstorm ideas for your own app

Here are some questions to help you think about what you might create. You might also want to use a generative AI tool for brainstorming.

  • What sorts of interactive elements have you seen in apps or on websites that intrigued you or that you think might have potential in your own work?
  • What types of stories do you usually tell or would you like to tell?
  • What form do those stories take? For example: written narratives, video news, sports profiles, restaurant reviews, trend or feature stories.
  • Think about how you usually tell those stories.  What additional elements might improve those stories or make them more engaging for your audience?
  • How might interactive elements in a web story or a standalone act allow you to engage with audiences in new ways?
  • What other tasks do you do in your journalistic work that might benefit from an interactive app?

Part 3: Create an interactive app

  • Once you have an idea, go to Google AI Studio.
  • Click on Build in the menu on the left. 
  • Enter your idea and generate the app.
  • Evaluate it and refine it through additional prompts. 
  • Also click on AI Features listed at the bottom left under Suggestions. That gives you the option to add image generation, audio tools, animation, maps, image analysis, and audio transcription, among other things. You don't have to use any of those, but seeing them might give you additional ideas. 

Part 4: Reflect on the process

Create a brief reflection that includes the following:

  • The name of your app and a link to it.
  • What did you create and why? How might it be used?
  • Do you see potential for tools like this in journalistic work? Why or why not?

Advanced AI Integration

If you are willing to invest some time into using Copilot, Gemini and other generative AI, you can create helpful interactive learning tools to use in your classes. None of these require expertise in coding. They do require expertise in subject matter and an ability to articulate directions and ideas into prompts and background information for generative AI apps. 

AI as a simulation tool

Chatbots can take on just about any persona you can imagine. You can instruct a bot on such things as the tone it takes with responses, the types of answers it provides, and the materials it draws on to provide answers. The output is never 100% reliable, but it is reliable enough for many types of class activities. Here are some examples:

  • Mock patient interviews. A physical therapy instructor at Michigan used generative AI to create a mock patient that students interviewed, allowing them to engage in a wide range of scenarios and to experiment with a variety of approaches they might use in the field.
  • Pharmacy coaches. At the University of Sydney, an AI agent provides feedback on students’ plans for managing patient medications. It also helps them prepare for national certifications. Instructors also created an agent to coach students on the financial aspects of pharmacy. That agent helped the instructor provide feedback and guided students on exam preparation. 
  • Occupational therapy simulator. At the University of Sydney, a Copilot agent simulates interaction with a busy kindergarten teacher. Occupational therapy students must interview the simulator, known as Mrs. S, to develop a support plan for a child. The agent is explicitly programmed to push back on suggestions, use jargon, and act impatient. This forces students to refine their skills in communication and negotiation.
  • PitchQuest: At the Wharton School at the University of Pennsylvania, faculty have created a simulator that students can use to practice pitching startup ideas. The simulator takes on investor personas that range from friendly to hostile and then offers detailed feedback on substance, strategy, and persuasion.

All of those were created with more sophisticated digital tools than we have access to at KU, but faculty can come close with Copilot. They would need to create a Copilot agent and give it instructions and material to draw from. CTE offers some resources to help: