Online Course Library

One scalable design system turned generic AI-generated courses into 200‑plus structured microlearnings.

AI makes production faster. It can also make problems scale just as fast.

An Australian training provider had used AI to generate 60-plus courses, and the same problems ran through almost all of them. Instead of scrapping them, I audited the library and wrote a three-phase plan the business approved. Then I led Phase 1: a new course design, a video template and a system the development team used to scale the fix.

Lead Instructional Designer · 200+ microlearnings · 100+ videos · $10,500 a year saved

See the before and after

A man taking an online course on a desktop computer, wearing headphones

At a glance

Client
An Australian registered training provider with two offers: formal qualifications, and a subscription library of microlearnings (short online courses of about 15 minutes)
Starting point
60-plus courses generated with AI before I joined, each based on an accredited unit (a nationally recognised unit from a formal qualification)
My role
Lead instructional designer. I audited the courses, wrote the plan, designed the new course structure and video template, and directed the development team that rolled it out.
Team
The provider's development team, trained by me and working from my templates. I reviewed their work against the template.
Tools
Coursebox (learning platform) · HeyGen (AI video) · Canva (design templates)

My Diagnosis

The courses looked complete. I audited them as a learner would. Because AI had generated them all the same way, the same four problems showed up in almost every one.

Weak instructional structure

Each course ran as one long block, with no consistent structure. Inside, lessons were long pages of unformatted AI-generated text with almost no interaction. They gave learners information, but nothing to do with it.

Poorly designed learning objectives

They described topics, not what the learner should be able to do, and were not linked to any lesson.

Flat avatar videos

A still photo with moving lips filled the whole video, with nothing on screen to support what was being said.

Third-party media

Generic AI-generated images from older models, mixed with third-party images, and YouTube videos doing much of the teaching. Nothing felt like the provider's own, and the provider had no control over the third-party media.

The challenge

Fix 60-plus courses quickly, without rewriting them from scratch.

A tight budget, no video studio, the provider's own developers doing the build, and a new brand rolling out at the same time.

That raised three design questions:

How do you make 60-plus long courses teach, without rewriting all the content?

Each course ran as one long block. The content itself was broadly right, but the form was not teaching. Rewriting 60-plus courses from scratch was not possible on this budget.

How do you keep quality consistent when a whole team builds the fix?

A development team would build them, not one designer. Quality could not depend on me checking every page by hand.

What should be standardised, and what should flex?

Structure, openings and intro videos had to look and work the same in every course. The content still had to fit each course's own objectives.

The fix had to reach every course, work within those limits, and use the content that already existed.

The design decisions

1. Restructure the courses into standalone microlearnings

One long block does not teach. I split each course into short, focused microlearnings that work alone and join into a learning path.

  • Standalone: each module became a microlearning of about 15 minutes, with its own goal, opening, intro video and short quiz
  • Connected: the microlearnings join into a learning path, a set sequence that rebuilds the full course
  • Two levels of assessment: the original quizzes and assignments, already aligned to the accredited unit, moved to the learning path. Each microlearning got its own short quiz and knowledge checks
  • Mapping kept: each microlearning kept its mapping to the accredited unit
  • Clear goals: objectives rewritten with measurable action verbs, and each one mapped to a lesson
  • Better lessons: the content was kept, not rewritten. Each lesson was reorganised into short chunks, with Coursebox interactions matched to each content type

Why this matters:

  • About 15 minutes suits a subscription library: short enough to finish in one sitting, focused enough to sell on its own.
  • Each course became 3 to 5 microlearnings that sell on their own, and together they still rebuild the full course: two products from one build.
  • Across 60-plus courses, that is 200-plus microlearnings.
  • A module cut out of a longer course doesn't explain itself. The new opening is what lets it stand alone.
  • Learners get a quick check in every short course, and the full course keeps the assessment already aligned to the accredited unit.

2. Phase the work: fix now, rebuild once for both offers

  • Phase 1: fix how every course is organised, and rebuild the lessons interactively, reusing the existing content
  • Phase 2: turn the accredited content, which existed as PDFs, into interactive modules the provider owns, built once for both offers
  • Phase 3: keep improving, guided by learner data

Why this matters:

  • Fixing structure didn't need new content, so it could be done across the whole library quickly.
  • Turning PDFs into interactive modules takes real design time for each unit, so it came second, once the foundations were right.
  • The same modules could later become a third product: the accredited unit, sold as an upgrade from the subscription.

3. Prove it on one course, then give the team a system

  • Prototype: I prototyped the new course structure on one course and the new intro video separately
  • Approval: stakeholders approved both before anything scaled
  • System: a course template, a step-by-step guide, writing prompts, a master video template and a thumbnail template
  • Rollout: I trained the developers and reviewed their work against the template

Why this matters:

  • Fixing a prototype is cheap. Fixing 200-plus microlearnings is not.
  • Templates improved consistency, and my review stayed part of production.
  • When the brand changed, it was updated once and rolled out everywhere.

4. Choose the video tool on what was needed, not what was planned

The provider had planned to use an enterprise avatar tool. I compared what the intro videos actually needed with the pricing, and recommended HeyGen instead. Then I designed one video template to replace the flat ones: an expressive presenter, supporting footage and motion graphics.

Why this matters:

  • A tool is paid for every year, so the right choice was the most affordable one that did the job.
PlannedRecommendedResult
Enterprise avatar toolHeyGenApproved by stakeholders

See the Work

The three-phase plan

The plan I wrote after the audit. The provider approved it. Select a phase to explore it.

Phase 1: Restructure into microlearnings

Goal: turn each long course into standalone microlearnings, using the content already built.

Key actions

  • Split each course into 3 to 5 microlearnings, one per module
  • Give each one its own title, description, goal, introduction and welcome video
  • Add four opening tabs: what you'll achieve, what you'll learn, how you'll apply it, and who it's for
  • Keep the existing lesson content and certification steps
  • Move the existing quizzes and assignments to the learning path, and add a short quiz and knowledge checks to each microlearning
  • Rebuild the lessons interactively: better formatting, Coursebox interactions matched to each content type, content organised and broken into chunks

Result

60-plus courses became 200-plus microlearnings without writing new lesson content.

What changed in every microlearning

Phase 1 kept the content and redesigned everything around it, including the lessons.

BeforeAfter Phase 1
Course structure
BeforeOne long course covering a whole unit
After3 to 5 microlearnings of about 15 minutes, each with the same structure: opening, lessons, assessment, what's next
Learning objectives
BeforeVague and inconsistent, like "gain a comprehensive understanding", with different lists on different pages
AfterWritten with measurable action verbs, and each one mapped to a lesson
Course opening
BeforeLong overview pages with repetitive and overlapping content
AfterOne clear opening: what you'll achieve, what you'll learn, how you'll apply it, and who it's for
Lessons
BeforeLong pages of article-style text, with little interaction
AfterContent kept, not rewritten. Reformatted, organised and broken into chunks, with Coursebox interactions matched to each content type.
Assessment
BeforeQuizzes and assignments at the end of one long course
AfterA short quiz and knowledge checks in every microlearning, and the original aligned quizzes and assignments on the full learning path
Intro video
BeforeA still photo with moving lips
AfterAn expressive presenter, supporting footage and motion graphics, from one template
Brand
BeforeMixed styles, on the old brand
AfterNew brand and one visual style across videos, thumbnails, banners and quizzes

The intro video, before and after

Two different courses: the Before video shows the old flat-avatar style, the After video shows the new template on another course. Press play to watch each one.

Before

Before: the original presenter video

  • A still photo with moving lips
  • The presenter fills the whole screen
  • Nothing on screen supports what's being said
After

After: built from the new template

  • An expressive presenter
  • Supporting footage and motion graphics
  • One template used across the whole library

What the project delivered

200+

microlearnings from 60-plus courses

Across Project Management, IT and Cyber Security.

15 min

per microlearning

Measurable objectives, chunked lessons, a short quiz and a designed intro video.

A system the team runs

Template, guide, prompts and video templates. The team built 100-plus intro videos from them, plus a new promotional video.

$10,500

saved a year

By choosing HeyGen over the planned enterprise avatar tool.

Want to upgrade your library at scale?

Available remotely for projects, contracts and roles.

Let's talkConnect on LinkedIn

[email protected]