It was the Sunday before the start of the semester and Priya Nair had a problem most instructors would recognize. She had a syllabus from the year before, a folder of PowerPoint slides and a D2L Brightspace shell that was, as she described it, embarrassingly empty. Priya wasn’t new to Brightspace. She had four years of experience with the LMS. She knew what she wanted: peer discussions at the end of each module, a short-graded quiz and a sequencing system that prevented students from skipping ahead. She had a clear vision. What she didn’t have was time. So, she did what a growing number of faculty do. She opened an AI tool, typed out her request in plain language and waited. What came back was, in a technical sense, impressive: A course design framework, learning outcome suggestions and a rationale for discussion-based learning. Everything was well-written, pedagogically sound — and almost completely useless for what she needed to do next: open Brightspace and build within the platform before her Monday morning class.
Pedagogy Is Not Configuration
The output Priya received may have been good course design, but what she needed was that design translated into Brightspace configuration. These aren’t the same thing, and the space between them is where most AI-assisted course design tools quietly falls apart.
Generic AI tools are trained on knowledge about pedagogy, learning theory and course design principles. They’re not trained on the specific tools, settings, workflows and sequencing rules of a particular platform. The result is advice that sounds right but doesn’t quite translate to things you can use without extra work.
Here’s what that gap looks like in practice:
| What you expect | What actually happens in Brightspace |
| Create discussions and quizzes, then connect them to grades | If grade categories aren’t set up first, each item lands uncategorized in the gradebook, defaulting to a weight of 10%. Three items silently claim 30% of the final grade. Brightspace flags an error and every item needs manual reassignment. |
| Set up grade items in advance, then build assessments | Forget to link the assessment to the existing grade item when building it, and Brightspace creates a duplicate automatically. The quiz points to the wrong one. The duplicate can’t be deleted because it’s attached to a live assessment. |
| Build a quiz inside an HTML Topic for instant student feedback | HTML Topics are display-only pages with no grade passback mechanism. Students complete the quiz, see their feedback and receive no grade. The gradebook stays empty. |
None of this is a flaw in Brightspace; they’re deliberate architectural choices in a mature platform. The problem is that knowing the correct sequence is exactly what four years of Brightspace experience quietly accumulates. General-purpose AI doesn’t have those four years.
Why Prompting Harder Doesn’t Fix It
When AI outputs miss the mark, the instinct is to prompt more carefully. Add more context. Be more specific.
This can help, but it doesn’t solve the core problem: the burden of knowing what to ask falls entirely on the instructor.
Priya would have needed to know, before she started, that she should ask about gradebook structure. She would have needed to know that the sequencing she wanted (modules unlocking in order) is called ‘Release Conditions’ in Brightspace, and that they’re applied on top of an existing module structure. The module plan must come first.
To get useful output from a generic AI tool for Brightspace course design, instructors need to know most of what you’re trying to figure out before they start. The faculty who need the most help are the ones least equipped to prompt their way out of this.
What’s missing isn’t more AI capability. It’s a layer that understands Brightspace well enough to translate a plain-language request into the right tools, in the right sequence, with prerequisite checks done before anything gets built or brought into Brightspace.
What That Looks Like in Practice
Consider how Priya’s request might go differently and imagine a scenario where she types the following:
Faculty: I need to redesign my intro course. I have a syllabus and some PowerPoint slides from last year. I want students to do a peer discussion and a short, graded quiz at the end of each module. And I don’t want them skipping ahead. They should have to finish one module before the next one unlocks. Where do I even start?
Orchestrator: Good news — you’ve described a clear redesign scenario, and this suite handles exactly what you need: module structure, discussions, quizzes and sequential unlocking. Before I map out your full skill path, a few quick questions. What’s the delivery mode?
Faculty: Blended. Are you working inside a Claude Project set up for this course?
Faculty: Yes, I have a Project for this course. What do you want to do with your existing syllabus and slides?
Faculty: Both — use the syllabus for structure and convert the slides. What’s the course code and name?
Faculty: SOCI 2210 — Social Inequality and Power. Roughly how many modules are you planning?
Faculty: Six to eight.
Orchestrator: Perfect. Here’s your full picture for SOCI 2210 — Social Inequality and Power.
Here are six to eight modules, each with interactive content pages converted from your slides, a graded peer discussion, a graded end-ofmodule quiz and sequential unlocking. That’s up to 24 assessment items plus module content pages. The suite can help map and prepare each piece of this and your syllabus gives us a strong starting point.
Order matters. There are two hard dependencies before you touch any assessments:
| Step | Skill | Why |
| 1 | redesign-planner | Maps your syllabus into a model structure; decides which pages need HTML Topics vs. standard tools. Everything else references this. |
| 2 | gradebook-planner | Must happen before Step 3. Grade categories, weights, and scheme must exist before any quiz or discussion item is created—otherwise nothing rolls up correctly. |
| 3 | slide-converter | Converts your PowerPoint slides into navigatable HTML Topic pages, one per module. |
| 4 | discussion-generator | Builds the peer discussion forum and topic settings for each module, with grading criteria. |
| 5 | rubric-builder | Builds the discussion rubric. Run alongside or just after Step 4. |
| 6 | quiz-generator | Generates end-of-module quiz questions and settings for each module. |
| 7 | release-condition-planner | Designs the unlock logic: quiz completion next module unlocks. Needs Steps 1–6 complete. |
| 8 | accessibility-auditor | Audits all HTML Topics produced in Step 3. |
| 9 | student-experience-preview | Walks through the course as a student — catches navigation gaps before students do. |
Upload your syllabus and run /redesign-planner to start. Your slides can wait until Step 3.
Faculty: Wait. Step 2 — gradebook? I didn’t say anything about a gradebook.
Orchestrator: You didn’t have to. Every graded assessment you build — quizzes and discussions included — needs a gradebook structure in place before it’s created. Without it, grade items end up in the wrong place, the final grade calculates incorrectly and fixing it is manual work. Two minutes now prevents an hour of cleanup later. That’s why it’s Step 2.
Priya mentioned slides, discussions, quizzes and sequencing. She said nothing about the gradebook. The flag wasn’t the result of better prompting. It came from a system that understood the relationships between Brightspace tools well enough to catch what was missing before any building began.
That’s the difference.
Not Automation. Translation.
A quick reading of that exchange might suggest AI is automating course design. It isn’t.
Priya’s instructional decisions are entirely hers: what to assess, how to sequence learning, what kind of discussion prompt generates genuine thinking. The AI made no pedagogical decisions. What it did was translate her vision into a Brightspace-ready workflow: the right tools, in the right order, with prerequisite checks done upfront.
In higher education, the AI conversation tends to circle around content generation, replacing human judgment and institution-wide rollout. It spends less time on the specific, unglamorous work of knowing how a particular platform is built and what it requires.
What Priya encountered wasn’t a smarter chatbot. It was a different architecture: one specialist skill for each task, and an orchestrator that knows which one to call, in what order and what to verify before it gets there. The AI capability is general. The workflow is Brightspace-specific. That distinction is what makes it useful.
The Question Worth Asking
General-purpose AI has made it easier to produce pedagogically reasonable course design output. But it hasn’t made it easier to turn that output into a well-structured Brightspace course, because that requires platform knowledge, prerequisite awareness and workflow sequencing that general-purpose tools weren’t designed to provide.
The gap between “what I want” and “what Brightspace needs” is still being crossed manually, by instructors with better things to do. Usually on the Sunday before the start of the semester.
The real question isn’t whether AI can help close that gap. We already know that it can when it’s purpose-built for the workflow. The real question is whether it can do so reliably throughout the full course lifecycle, for instructors at every experience level, without requiring them to already know the answers.
In Part 2, I’ll introduce the Brightspace Interactive Design Skill Suite: 40 specialist skills and the orchestrator that routes between them. I’ll show what that looks like in practice.
Author’s note: This post was drafted with the assistance of an AI writing tool. All ideas, arguments, examples and conclusions are my own. I reviewed, revised and take full responsibility for the content as published.
This post is Part 1 of a three-part series on AI-assisted course design in Brightspace. The Brightspace Interactive Design Skill Suite is licensed under CC BY-NC 4.0.
Kumar Chandrasekhar, PhD · Mount Royal University · CC BY-NC 4.0
Written by:
