After 11 years in this industry—from wrangling LMS bugs to staring at thousands of lines of Storyline variables—I’ve learned one thing: no matter how fancy the technology gets, the "delete" key is still the most powerful tool in an instructional designer’s arsenal. For the past 18 months, I’ve been heavily integrating AI into my workflow. It’s a productivity superpower, sure, but it’s also a high-speed engine for creating well-formatted, grammatically correct nonsense.
When we use AI to draft storyboards, the role of the instructional designer shifts from "writer" to "editor-in-chief and auditor." If you think you can skip the QA phase because the AI sounded confident, you’re setting yourself up for a mess in the LMS. Here is how I approach a storyboard QA checklist, ensuring that our AI-assisted work is actually learning-ready.
The Philosophy: What "Validation" Means in an AI-Assisted World
Validation used to mean "did the SME agree with this?" Now, validation means "did I ensure this machine didn't hallucinate a process or invent a policy that doesn't exist?" When we use AI to draft, we aren't just checking for typos; we are performing a structural audit. My running "gotchas" doc is filled with examples of AI inserting subtle contradictions—using a specific term on page one and its synonym on page five, or suggesting a quiz question that ignores the context provided in the slide. Validation is about maintaining the integrity of the instructional design, not just the grammar.
Risk-Based QA: Triage Your Review
Not every course requires the same level of scrutiny. If you treat a "How to use the coffee machine" deck with the same intensity as "Sexual Harassment Prevention Training," you will burn out your SMEs. I use a risk-based matrix to decide how hard I’m going to poke at the AI-generated content.
The Core eLearning Review Checklist
When I review a storyboard, I’m looking for the "AI fingerprint"—that overly formal, slightly repetitive, and often vague corporate voice that makes learners' eyes glaze over. Here is the instructional design QA checklist I apply to every AI-generated document.
1. Structural and Alignment Checks
- Learning Objective Alignment: Does every single interaction or slide content clearly map back to an objective, or is the AI just "fluffing" the word count? Logical Flow: Did the AI skip a step in a process? AI loves to assume the learner knows the intermediate steps that were mentioned in a previous version of a policy. Cognitive Load: Is the text on the slide "textbook heavy"? If the AI provided a paragraph, split it into three bullet points or move it to a "Click to Reveal."
2. Content Consistency Checks
- Terminology Audit: Search and replace the AI’s favorite filler words (e.g., "leveraging," "robust," "seamlessly"). Ensure specific internal terms are used consistently throughout the storyboard. Voice Consistency: AI often shifts tone. Check if it started with a friendly, conversational "we" and ended in a cold, passive-voice "the user must." The "Gotcha" Look-up: Scan for contradicting statements. I once had an AI suggest a "Submit" button in one slide and an "Auto-advance" in the next.
Fact-Checking and Source Tracking
This is where I get pedantic. Never take an AI’s word for it. If the storyboard says, "Company policy states X," I demand that the AI provides the source document or I verify it against my own local file. Instructional design QA is failing if you aren't verifying the accuracy of the claims.
I maintain a strict rule: if the AI generates a statistic or a quote, I tag it in my storyboard document with a placeholder like [SOURCE: VERIFY]. I won't move to the development phase until that tag is removed and replaced with a hyperlink to the internal Wiki or source document. If the AI can’t tell me where it pulled the info from, I treat it as a hallucination until proven otherwise.
Breaking the Assessment: My Favorite Part
Nothing reveals the limitations of an AI-generated storyboard like a poorly constructed assessment. AI is notoriously bad at creating "distractors." It tends to make the correct answer obvious or makes the distractors so absurd that they aren't even plausible.
When I test assessment questions as a learner, I try to break them:
The "Technically Correct" Test: Can I choose a wrong answer that is technically correct based on a vague sentence in the course? The "Absolute" Trap: AI often uses words like "Always" or "Never" in questions. I rewrite these to be more realistic and nuanced. The "Stem" Ambiguity: I rewrite the question stem until it is impossible to misinterpret. If I find myself needing to explain the question, the question is broken.Efficient SME Review: How to Not Waste Their Time
Don't send a raw AI draft to a SME. That is the quickest way to lose their AI governance framework for instructional design respect. I perform a "first pass" myself to remove the SME review workflow training AI fluff and polish the tone. By the time the SME sees the document, the instructional design is already sound.
Strategies for a Targeted Review:
- The "Ask-The-Question" Method: Instead of "What do you think of this slide?", use "Does this scenario accurately reflect the common complaints you receive from customers?" The "Highlighted" Focus: Use comments to highlight the areas where you specifically used AI and need the SME to verify the *facts* rather than the *writing style*. SME Timeboxing: Be clear about what you need. "I need you to verify the steps on slide 4 and 5. The rest is standard compliance fluff."
The "Reviewer’s Mindset" Manifesto
When you are performing your elearning review checklist, avoid the trap of "looks good to me." That phrase is a death sentence for quality. If you don't find at least three things to change in an AI-generated storyboard, you aren't looking closely enough.
AI is a draft engine, not a final product. We are the architects. Our value isn't in how quickly we can churn out slides; it’s in our ability to catch the nuances, the tone, and the factual errors that a machine simply cannot perceive. Keep your "gotchas" doc updated, stay cynical, and keep questioning the machine. Your learners will thank you for it.
Looking for more tips on building a robust L&D workflow? Check out my other guides on LMS administration and effective project scoping. And remember: if you don’t test your own assessment questions until you break them, you aren't finished yet.