Background
Overview & Context
Designing technically rigorous online courses at scale requires balancing speed with quality. At WGU, course development involves coordinating with subject matter experts who bring deep technical knowledge but limited time — extracting that expertise, translating it into structured learning objectives, and building assessments that accurately reflect real-world practice is time-intensive work, often bottlenecked by back-and-forth revision cycles.
As the volume and complexity of course development projects grew across the College of IT, I began looking for ways to make the design process more efficient without sacrificing the instructional integrity that competency-based learning demands.
Problem Statement
The Challenge
The core bottleneck wasn't a lack of subject matter expertise — WGU's SMEs are deeply knowledgeable practitioners. The problem was a structural one: the traditional ID workflow required SMEs to function as authors, producing written content from scratch, which is not where their time or skill is best spent.
The drafting and iteration phases of course development were consuming disproportionate time — not because the content was unclear, but because the process for capturing and structuring it was inefficient.
Every new course project restarted this same friction: lengthy SME interviews, rough notes converted manually into learning objectives, multiple revision rounds, and late-stage misalignments between objectives and assessment items. A more systematic approach was needed — one that could scale across a growing portfolio of courses without a proportional increase in development hours.
Design Process
Approach & Methodology
What started as experimentation with AI tools on a single course project evolved into a structured, repeatable prompt-engineering workflow embedded throughout the full instructional design process. Rather than treating AI as a general-purpose assistant, I focused on the specific friction points in the ID workflow and built purpose-designed prompts for each.
Identifying Friction Points
Analyzed the existing ID workflow to locate where time was being lost — specifically the phases where SME input was being manually transformed into structured instructional content. LO drafting, content structuring, and objective-to-assessment alignment emerged as the highest-impact targets.
Learning Objective Development
Built prompts that could take raw SME input — notes, outlines, rough content drafts — and generate structured, Bloom's-aligned learning objective candidates for review and refinement. This shifted the work from generation to evaluation, significantly compressing the early design phase.
AI-Assisted Content Drafting
AI-assisted drafting allowed production of structured first-draft content from SME source material. SMEs could then review and correct rather than author from scratch. For technical SMEs whose bottleneck was writing time rather than knowledge, this fundamentally changed the collaboration dynamic.
Reusable Prompt Template Library
As the workflow matured, the most effective prompts were documented and systematized into a reusable template library. This created a repeatable process applicable consistently across courses — and shareable with team members — rather than a one-off approach rediscovered on each project.
Objective-to-Assessment Alignment Check
Applied AI prompts to cross-check alignment between learning objectives and assessment items, catching gaps earlier in the process and reducing the late-stage rework that had previously extended development timelines.
Tools used across this workflow:
Artifact
From Raw SME Input to Bloom's-Aligned Objectives
The following example illustrates the core workflow in action. The input below is an unedited excerpt from a subject matter expert describing what learners should know about AI prompt construction — captured in a working session, not polished for documentation. The output shows the Bloom's-aligned learning objectives generated from that input using a structured prompt, ready for instructional designer review.
"Ok so basically what I want people to know is that when they're writing prompts for AI they need to think about what they're actually asking. Like a lot of people just type something vague and then complain the AI doesn't give them what they want. The main thing is you have to give it context. Also telling it who it is helps — like say it's a project manager or whatever role fits. And you need to say what format you want back otherwise it just gives you a wall of text. I've seen people ask it to 'write something about the project' and then wonder why it's not useful. Oh and the output — make sure you tell it if you want bullet points or an email or a table or whatever. I guess the other thing is don't be vague, like 'help me with this' is useless. You need a specific task. I think that covers the main stuff. Maybe also say something about how being specific saves time because you don't have to keep asking it to redo things."
By the end of this module, learners will be able to:
-
1Explain why vague prompts produce inconsistent AI outputs by describing how specificity, context, and role assignment improve AI-generated results. Source: "a lot of people just type something vague... you have to give it context"
-
2Apply the Role + Task + Context + Output Format framework to construct a complete AI prompt for a workplace task. Source: "telling it who it is helps... give it context... say what format you want"
-
3Distinguish between weak and strong prompts by evaluating examples against the four-component framework. Source: "'write something about the project'... 'help me with this' is useless"
-
4Select an appropriate output format — such as bullet points, email, or table — based on the intended use of the AI-generated content. Source: "tell it if you want bullet points or an email or a table"
-
5Demonstrate how writing a specific, well-structured prompt reduces iteration time and improves the quality of AI outputs on the first attempt. Source: "being specific saves time because you don't have to keep asking it to redo things"
Objectives follow Bloom's Taxonomy action verbs: explain, apply, distinguish, select, demonstrate.
Results
Outcomes & Impact
Applying this AI-integrated workflow across courses in the College of IT produced measurable improvements across several dimensions of the development process.
Beyond raw efficiency, the workflow produced qualitative improvements that are harder to quantify but equally significant. Structured prompt outputs created a more uniform baseline for content and objectives across courses, reducing the variability that typically comes from juggling multiple projects and SMEs simultaneously. Objective-to-assessment alignment gaps — previously caught late, after significant rework — were surfaced earlier in the design phase.
Perhaps most meaningfully: shifting SMEs from authors to reviewers changed the nature of the collaboration. Giving SMEs AI-assisted drafts to react to rather than blank pages to fill shortened review cycles and improved the quality of their feedback — because evaluating a concrete draft is cognitively easier than generating content from scratch.
Reflection
What I Learned
The most important lesson from this project isn't about AI — it's about workflow design. The prompts themselves were only effective because they were built around a clear understanding of where friction existed in the instructional design process. Without that analysis, AI assistance would have been applied randomly rather than strategically.
The shift from generation to evaluation — having both SMEs and designers spend their time refining rather than producing from scratch — turns out to be a broadly applicable principle. It's not specific to AI; it's a sound design pattern that AI simply makes more accessible at scale.
The reusable template library was the highest-leverage output of the project. A workflow improvement that lives in one person's practice is fragile; a documented, shareable library has institutional staying power. If I were starting again, I would have prioritized the documentation phase earlier rather than waiting until the workflow was fully mature.