Designing a module with AI: How ChatGPT streamlined our development process

Photo by charlesdeluvio on Unsplash

Generative AI can streamline the design of online modules, freeing academics for high-value tasks while learning designers shape content. This post shares a step-by-step approach and key insights on balancing AI with human expertise.

---

As a learning designer I have been exploring how generative AI can support the development of new online courses. In this example I’ll discuss how I used ChatGPT to develop a new online postgraduate module we’re developing with Royal Holloway, University of London, and how it enabled us to get the best value from our academics’ limited time.

Before we dive in, it’s important to point out that simply relying on ChatGPT to create a well-designed course is likely to be disappointing. However, if you’re prepared to do some work in partnership with the tool, the results can be pretty good.  

This is how I approached the task, and what I discovered:

1. Gather the basic facts about the module

We work in partnership with academics from our federation members so, to get started, I needed some basics details about the module from them:

• Module learning outcomes
• Tiles for each week of study
• Learning outcomes for each week of study
• A reading list

These foundational elements provided the framework around which I built the module.

2. Generate an outline of the module content

I started by using ChatGPT to outline the content of the module such as learning activities, reading materials and other resources. To make this manageable, I divided the task into small chunks and worked on a week of study at a time.  I provided focused prompts which included details such as:

Programme name	MSc Project Management with Software Development
Module name	Information Security and Governance
Academic level	Postgraduate
Mode of attendance	Online, asynchronous study
Learning outcomes	From module specification
Structure	Three lessons per week
Types of activity	Readings, videos, case studies, etc.

Here’s an example of my initial ChatGPT prompt:

3. Interrogate the response, start to build the structure

With the information I provided, ChatGPT produced a plan which turned out to be a useful starting point. Here’s what it produced based on my initial prompt:

This wasn’t exactly what I needed. I found that I had to take this response and iterate – it’s never right in the first try. Instead, I treated it like a rough brainstorm, refining and reshaping it repeatedly to ensure depth, clarity, and proper alignment with learning outcomes.

In this example, I reviewed and refined its suggestions by:

• Picking out the bits of its suggested structure that I liked and drilling down into more detail
• Asking clarifying questions to ensure conceptual accuracy
• Drawing links between concepts to get the structure right
• Ensuring alignment with the reading materials provided by the academic
• Incorporating best practices from my learning design knowledge and our university’s learning design framework, the CAFÉ model (Content, Activity, Feedback, Evaluation)

The process felt similar to running an in-person learning design workshop, with ChatGPT taking the role of an ever-present assistant to brainstorm, refine, and adapt content.

4. Generate full activities, more iteration and iteration again!

Next, I asked ChatGPT to expand on some of the activities so I could see what they would look like. For me, step 3 and step 4 aren’t strictly linear – I don’t necessarily complete the outline before moving on to generating full activities. Instead, I go back and forth between refining the outline and developing activities, adjusting as needed.

I often found the activities ChatGPT generates to be too simplistic or lacking sufficient context. For example, when generating case studies, I noticed that it frequently left out critical details students would need to answer questions effectively.

Here are some examples of further prompts I used to refine the content:

By continuously prompting the AI to refine activities, I ensured they were challenging and appropriate for postgraduate students, well-outlined with clear instructions and aligned with the learning objectives and the CAFÉ model.

I also found it helpful at this stage to start building the course in our Word templates as I went, so I could fully visualise the module and the gaps I had to fill.

5. The role of the academic

Together, ChatGPT and I had designed a lot of the module which preserved the precious academic time for some really valuable tasks:

Write and present lectures

One area that I think AI falls short is in writing lectures; I found it struggled to create the engaging, personal tone we wanted. Human expertise is essential for creating engaging, high-quality lectures. Having an academic on screen, explaining ideas in their own words, brings credibility and a human touch that AI simply can’t replicate.

So, our academic focussed on writing their lecture scripts and taking the time to come into our studio to film, while I worked on developing the activities.

Review the module design

Once I completed a draft of the module, I shared it with the academic to review. Their feedback focused on:

• Alignment with module and weekly learning outcomes
• Appropriateness for postgraduate-level study
• Accuracy and depth of content
• Structure, flow and scaffolding

What was the academic’s reaction to the AI generated content?

Generally, we have found that academics are impressed with the output we get from ChatGPT, but that it often needs refinement. In this example, our academic identified that the focus on enterprise architecture frameworks was too detailed for the programme (it wasn’t an IT programme, it was for project managers in the tech space). This is the sort of nuance that AI struggles with and we really need to rely on our academics for.

Our academic’s input led to further refinements, ensuring academic rigour while maintaining efficiency in the development process.

How did it go?

Here’s a summary of what went well, and what we can do to improve the process in future.

Successes

1. Ensures academic time is spent well: AI accelerated module development by shifting the role of the academic from content creator to content reviewer and enabling them to focus on high-value tasks such as making video lectures and quality assurance. It significantly reduced their workload while maintaining academic integrity and quality.
2. Production becomes more predictable and reliable:  Working this way gave more control to our learning designer in the production process, which is advantageous to us as we have more direct control and accountability over the learning designer’s workload than we do over the academic’s.
3. Quality improved: For academics new to online learning, AI helped ease the learning curve by enabling them to focus on content and quality, rather than the complexities and pedagogies of online teaching.

Key takeaways

1. Obtain more information upfront: In the future, I would ask our academic to write their lecture scripts before I begin the work with AI. I could then feed this into ChatGPT, providing a richer knowledge base and clearer direction, and ensuring better alignment with learning outcomes and key concepts.
2. The human element is essential: The academic’s role is crucial – they are needed for the essential tasks of sense checking content, ensuring accuracy, and infusing personal expertise into video lecture.
3. Expect to work with ChatGPT to get a good result: There’s a lot of emphasis on ‘prompt engineering’ in discussions around genAI. While it’s important to provide ChatGPT with sufficient information, I don’t believe there’s a single perfect prompt that will always deliver the exact answer you want. The focus should be on iteration and having the confidence to ask questions and suggest adjustments until you’re satisfied with the result.
4. Develop a curated GPT model: I’m interested in exploring a curated GPT for module development. By inputting examples of our best content, the CAFÉ model, preferred structures, and other relevant programme details, we could help the GPT better understand the module’s background, reducing the need for constant reminders.

What does this mean for the learning design role?

It’s natural to worry about how new technologies – especially something as transformative as AI – will impact jobs. Personally, I enjoyed using AI to design this module. It gave me more control over the final content and encouraged deeper creative and critical thinking as I evaluated what was generated, how it fitted within the module, and how it aligned with the learning outcomes. While I previously suggested that AI shifts academics from creators to reviewers, I found the opposite to be true for learning designers – it transformed my role from reviewer to creator.

That said, although it reduces the academic’s time, this process requires a significant time investment from the learning designer. Some may hope that AI simply generates usable content from a single prompt but, in my experience, that’s not the case. If learning designers and academics were expected to take that approach, the work could become very unfulfilling.

Ultimately, AI is a tool that, when used thoughtfully, can enhance learning design rather than replace it. The key is knowing how to work with AI effectively, and this is an exciting space for learning designers to explore.