Turn Word into Moodle books
A new AI assistant developed at UNE Partnerships, addresses one of the most time-consuming aspects of creating online learning materials - converting Word documents into clean HTML for Moodle books.
The Word to HTML assistant tackles a workflow challenge familiar to many educators: transforming lengthy course documents into properly formatted, web-ready content.
Converting content, not just copying it
UNE Partnerships works with vocational education and training providers to create learning and master resources - comprehensive Word documents developed with subject matter experts that form the foundation of online courses.
"We get units of competency and we create what we call learning master resources, which is this big document in Word," explains assistant creator Samir Bernardes. "We go through the content with subject matter experts and we create the chapters which then become Moodle books for students to use in our online course."
The traditional process of converting these documents to HTML has been labour-intensive. Word's native HTML output can be problematic, often producing bloated code with unnecessary formatting tags that cause rendering issues in Moodle. Manual cleanup is typically required, making the process tedious and inefficient.
Teaching the assistant UNE's code template
Working with Shannon Tyrrell (AI Innovation Lead at LabNext70), Samir created an assistant instructed to generate HTML based on UNE's specific code template - ensuring consistency in headings, colours, and activity layouts across all course materials.
"I taught my assistant my code template and it converted into HTML, which you can then copy directly into your Moodle book," Samir demonstrates.
The assistant handles the bulk of the HTML coding automatically, including complex elements like tables. Rather than attempting to transfer images directly - which often causes problems in Word-to-Moodle conversions - the assistant inserts placeholder markers (asterisks) wherever an image or link should appear.
"I asked it to just give me asterisks as a placeholder so I can easily go there and upload the link that I want, so I ensure that the link is correct," Samir explains. "That way we verify the link, we verify the accuracy, and if the link is up to date to the content."
Human oversight remains essential
The approach reflects a practical understanding of AI's role in content creation. While the assistant handles the time-consuming HTML coding, staff retain responsibility for verification and quality control.
"You still have to go through and filter," Samir notes. "Basically I check the information, and if everything is there... I still need to tweak bits and pieces, but the heavy bulk HTML coding is done."
The assistant uses Gemini 2.5 Flash, though Samir notes that Claude 4.5 Sonnet also performs well for this type of task. The instructions evolved through iterative refinement, with successful approaches from initial conversations being incorporated into the assistant's core instructions.
Broader applications
While developed specifically for UNE Partnerships' vocational education workflow, the assistant has potential applications for any staff creating Moodle books. The time savings allow course developers to focus on instructional design and content quality rather than manual HTML coding.
"This can be helpful for people that are not really into HTML and create Moodle books and so forth," Samir suggests. "You can just design everything in Word and as long as you are aware of the code template rules and definitions, it will just code everything for you. Now that we are saving time with manual html conversion, our final goal is to improve the MVP by creating a new code template that is more engaging and interactive and streamline the process towards a better product."