How AI-Powered Notebooks Are Changing Courseware Development

Courseware and training is an essential part of customer success in enterprise software. With the rise of AI tools, new opportunities have emerged to streamline the creation of relevant, high-quality training materials. One such tool is NotebookLM, an AI-powered notebook developed by Google. This blog explores how NotebookLM fits into the courseware development process at Progress, particularly within the ADDIE framework, and evaluates its potential to transform the way we design and deliver educational content.

The Courseware Developer Role

Within software companies, courseware development is sometimes viewed as a specialized, supplemental extension to technical writing. While a technical writer provides comprehensive documentation detailing every product feature and function, a courseware developer focuses on guiding users through the best practices for achieving specific goals with the product.

While there can be many ways to accomplish a software task, the courseware developer's job is to speak on behalf of product experts to guide users through the "happy path" of the tool. The role is more opinionated than that of a technical writer because a courseware developer must be able to defend their decision to choose one path over others.

Courseware development also differs from technical writing in that it requires a dynamic mix of instructional elements (videos, interactivities, graphics, quizzes) in addition to the text-based elements found in technical writing.

At Progress, we follow a well-known framework for courseware development called the ADDIE Model of Instructional Design. We also use the scaled agile framework (SAFe) to plan our quarterly work using story estimation points.

In the ADDIE Model, education development has five phases (with broad steps described):

• 🔍 Analysis (3-4 story points): Define project requirements, gather resources and establish course goals through a Course Plan document. Validate with subject matter experts (SMEs). 
• 📐 Design (5 story points): Define the specific lessons, topics and instructional elements in the course through a Detailed Course Outline document. Obtain approval from SMEs. 
• 🏗️ Development (10-15 story points): Create the content for each lesson (text, image, audio, video, and other interactive multimedia), post the self-contained lesson to the sandbox website, and have content reviewed by peers and SMEs.
  • The Development work is approximately 50% writing content (including video scripts) and 50% using other tools to create video, audio, interactive or graphical content and review content with SMEs. 
• 🚀 Implementation (Posting) (2-3 story points): Post to live website, archive course files, announce course availability, and promote on product communities. 
• ✅ Evaluation (1-2 story points): Update lessons as required, track learner metrics and feedback, and provide course completion data to stakeholders.

AI Tooling Landscape for Courseware Development

Since the proliferation of LLM-based AI technologies in 2023, small vendors in the eLearning space have made claims such that their tools can generate a complete course from a single PowerPoint deck. In 2024, we sampled some of these tools at Progress and found they fell short of their promises. We also found the licensing expensive and service-level objectives opaque (one cloud-based tool was offline all morning with no messages on the company's website or social media channels).

If 2023 and 2024 were the years of proliferation, 2025 has been the year of consolidation. The major vendors have been deploying and promoting features that the niche players offer as entire products and the new features are often included in licenses we already have.

This year, courseware developers at Progress have built custom GPTs using ChatGPT and Microsoft Copilot. These tools have proven helpful in gathering source content from the product documentation and adhering to the corporate style guide. However, each tool limits the number of files it can reference and each can pull data from irrelevant or imagined sources.

When Progress's Chief AI Officer heard of our experiments with custom GPTs, he suggested we try a different type of tool, an AI-powered notebook, whose defining characteristic is that it only answers questions about the data you provide.

What Is NotebookLM?

NotebookLM is an AI-powered research and writing tool developed by Google that acts as a virtual research assistant. Unlike general AI chatbots that pull information from the internet, NotebookLM focuses solely on the content you provide. This Resource Augmented Generation (RAG) approach helps to minimize hallucinations and verify the AI's responses against your uploaded sources.

You upload various types of files to NotebookLM, including:

• PDFs
• Google Docs and Slides
• Copied text
• Website URLs
• YouTube videos (it extracts the transcript)
• Audio files (it transcribes them)

These files become the Sources for your Notebook. After uploading your sources, you can query NotebookLM using prompts, similarly to other chatbots.

A unique feature of NotebookLM (even among other AI Research Assistant tools) is that it can turn your documents into engaging, two-host podcast-style discussions, allowing you to listen to your research.

How Does NotebookLM Differ from Other LLMs Like Copilot, Gemini and ChatGPT?

Both technologies make use of LLMs to process and generate text. Users interact with both technologies using prompts but the purpose and functionality of AI-powered research and writing tools is fundamentally different than general-purpose AI-powered chatbots.

NotebookLM is a specialized AI application focused on knowledge analysis and content generation. It acts like an intelligent research assistant dedicated to your specific knowledge base.

There are at least two functional differences between the technologies that become apparent quickly:

• NotebookLM will try hard not to answer general questions. It only wants to pull responses from your sources.
  • Let’s say you ask it, “How do I bake a cake?” Unless your sources contain recipes, NotebookLM’s answer will likely be: “Based on the sources provided, there is no information available on how to bake a cake.”
• Prompts are not saved in NotebookLM, although Responses can be saved as persistent Notes and converted to Sources. You can’t count on your chat history being available in later sessions.

Notebook LM for Courseware Development at Progress

NotebookLM has been most helpful in the Design and Development phases of the courseware creation process. This technology should prompt a reevaluation of activities in the courseware development process.

Here are the efficiency gains I’ve found in each phase and suggestions for incorporating NotebookLM into the phases of the ADDIE model:

🔍 Analysis – Small Gains

This phase is primarily about stakeholder communication and resource gathering.

Suggested use of NotebookLM:

• Set up a course notebook early and upload all relevant sources, including the Course Outline.
• Use prompts to explore the Course Outline (e.g., “What do you know about feature X?”) to familiarize yourself with the tool.

📐 Design – Large Gains

After uploading sources, NotebookLM can assist in drafting the Detailed Course Outline (the course blueprint). While not perfect, these drafts provided a strong starting point to bring to SME discussions and refinement.

Efficiency improvements:

• Reduced effort from ~5 story points to ~1–2 points.

Suggested use of NotebookLM:

• Generate course structure, including: lesson breakdowns, opportunities for videos or guided exercises, topic-level suggestions.

🏗 Develop – Large Gains

NotebookLM was effective in generating first drafts for lesson content, video scripts, and guided exercises. It was nearly 100% effective in generating lesson summaries and “Check Your Understanding” questions.

Efficiency improvements:

• Reduced text-based content creation from ~5–7.5 story points to ~1–3 points.
• Therefore, the total effort in this phase went from ~10-15 points to ~6-10 points.

Suggested use of NotebookLM:

• Generate first-draft lesson content based on course blueprint.

Notes:

• Review and refinement time increased.
• Non-text content (e.g., images, videos) was unaffected, as it relies on other tools.

🚀 Implementation (Posting) – No Gains

This phase involves posting to systems like Sitefinity, Progress Community and SharePoint.

NotebookLM has no current role due to lack of integration with these platforms.

✅ Evaluation Phase – No Gains (Yet)

This phase is already largely automated via reports and scripts.

Future potential use case: Provide NotebookLM updated release features and existing course content to suggest which courses may need updates.

Limitations of NotebookLM

In my testing, I found the following limitations when working with NotebookLM:

• NotebookLM does not allow Microsoft Word (DOCX), PowerPoint (PPTX), or Excel (XLSX) file types as sources.
  • Microsoft is rolling out its own AI-powered notebook technology (Microsoft 365 Copilot Notebooks), which may ultimately be a better provider of this technology.
• The gains in efficiency are dependent on having plentiful source content.
  • If feature discussions take place via Teams or Slack (or water coolers), it might be difficult to gather this information, as there is no native integration in place for those channels.
• This tool has no effect on the efficiency of producing images or videos (except that it helps to write the script). Other tools are needed for these types of content.

Evolving Skills for Courseware Developers

As NotebookLM streamlines content generation, the role of courseware developers is shifting from content creation to content curation, refinement and strategic design.

I see the following skills as becoming increasingly important:

🧭 Information Sourcing and Organizational Awareness

• Identify and gather relevant content from numerous sources—including people.
• Evaluate metadata to determine content freshness, accuracy, and relevance. 

🧩 Instructional Architecture and Systems Thinking

• Design logical course flows that build progressively from topic to topic.
• Adapt course sequencing as new ideas or requirements emerge during development. 

👩‍🏫 Subject-Matter Expertise and Content Validation

• Verify that AI-generated content aligns with product knowledge and stakeholder expectations.
• Confidently represent the material in SME reviews and public-facing content. 

✒️ Editorial Precision and Style Consistency

• Refine AI drafts to meet the company style guide and maintain a consistent instructional voice.
• Apply strong proofreading and editing skills to elevate clarity and tone. 

🎨 Information Design and Learner Engagement

• Use saved time to enhance how content is presented—visually and interactively.
• Select the right instructional elements (e.g., videos, guided exercises, simulations) to maintain learner interest and reinforce understanding. Become skilled at creating these types of content. 

📈 Continuous Improvement and Trend Awareness

• Monitor evolving e-learning trends and learner preferences.
• Gather feedback and experiment with new formats to improve course effectiveness.

Conclusion

NotebookLM represents a promising shift in how courseware developers approach content creation—moving from manual drafting to strategic curation and refinement. While it doesn’t replace the need for subject-matter expertise or instructional design skills, it significantly accelerates early-stage development and supports more informed decision-making. 

As AI tools like NotebookLM continue to evolve, they will likely become indispensable collaborators in the courseware development process, enabling courseware developers to focus more on learner engagement, innovation and continuous improvement.