Same Prompt, Different Tools: Mindsmith
I've been playing around with a lot of AI tools lately. There are so many of them, and they all seem to approach the work a little differently. That made me curious about what would happen if I started them from the same place and let each tool make its own decisions.
I'm giving each tool the same problem, the same source material, and the same basic requirements. I'm not telling them what kind of course to build or what instructional strategies to use. I want to see what decisions they make on their own.
What instructional choices does the AI make? Does it create meaningful practice or just present information followed by a quiz? What does it decide is important enough to include? And, especially important to me, how accessible is the result?
I look at accessibility from two angles: how accessible the content and experience is “out of the box,” and whether I’m able to make the changes needed to make it accessible within the tool.
The Prompt
I gave each tool the same general prompt and source material. Keep in mind, this is skipping most of the instructional design process. This is all hypothetical so I could keep the scope fairly narrow. I found this OSHA document on Heat Illness Prevention, and decided to narrow it down to landscaping-related work environments.
Here is the exact prompt that I pasted into each tool:
"Create a 10–15 minute digital learning experience for newly hired seasonal employees at a landscaping company who will be working outdoors in hot weather.
Employees receive information about heat safety, but supervisors have found that new employees do not consistently recognize signs of heat illness or know how to respond appropriately when they or a coworker begins experiencing symptoms.
Use the provided OSHA document, Protecting Workers from the Effects of Heat, as the source material. The learning experience should prepare employees to apply this information and make appropriate decisions when they encounter potential heat illness on the job.
The learning experience must conform to WCAG 2.2 Level AA.
Use a clean, professional visual style appropriate for the audience, with a color palette inspired by the provided OSHA document."
I didn't give any more instruction unless I was specifically asked questions that I needed to answer in order for the tool to move forward.
This is only the first in a series of experimenting with different tools, so follow along for the fun!
First up: Mindsmith.
About Mindsmith
Mindsmith is an AI-native eLearning authoring tool. This is exactly what it's built for.
They offer a free plan if you want to check it out for yourself. I have the Professional plan, which you can currently purchase for $468 per year or $49 per month. (I love a monthly plan option!)
One thing that stood out to me right away was the support. I asked the AI chatbot a question about credits, and when it gave me a not-completely-correct answer, Zachary (a human!) jumped into the chat and immediately gave me the correct information. I was pleased to see that the people behind an AI-native tool are still keeping a close eye on things.
Also, HUGE props to Mindsmith for how they responded to my accessibility feedback. As soon as I finished evaluating the output, I sent a lengthy email detailing the accessibility issues I encountered and wasn't able to fix within the tool. No more than 15 minutes later, on a Saturday night, I had a response thanking me for reaching out and letting me know that my findings were being forwarded to their accessibility engineer.
They also shared that they're working with a professional accessibility auditor to identify and remediate issues, with the goal of repeating that process at least annually. I'm looking forward to seeing the improvements they make!
How Mindsmith handled the prompt
After receiving the prompt, Mindsmith asked me a series of follow-up questions. Here are the questions it asked and the answers I selected:
Q: What visual style should the images have?
A: (skipped)
Q: The OSHA fact sheet is fairly dense and technical (12-element prevention program, indoor heat stress controls, etc.). Since your employees already got the baseline heat safety info and the gap is specifically recognizing symptoms and responding, should I pull out and focus only on the recognition/response content (signs, risk factors, emergency actions) and treat the broader prevention-program material as background I don't need to fully cover, or do you want the full document represented?
A: Focus tightly on recognizing symptoms & responding correctly
Q: Supervisors flagged that new hires struggle to recognize symptoms and respond correctly, in themselves or a coworker. Should the core practice activity be a branching decision scenario (e.g., a coworker starts showing symptoms mid-shift, what do you do) as the centerpiece?
A: Yes, make a branching scenario the centerpiece
Q: Any company name, logo, or brand colors I should work into the lesson, or should I just build a professional palette inspired by the OSHA document's colors (e.g., its blues/oranges)?
A: Just use OSHA-inspired professional palette, no branding
Good — this gives me solid material. Before I lock in a plan, let me check in with you since imagery was skipped and I want to confirm my read on structure.
Q: Since you skipped the imagery question, my plan is: clean flat-style illustration (not photo-realistic, no stock-photo feel) in an OSHA-inspired navy/blue + safety-orange palette, used purposefully for pacing and a couple of teaching diagrams (not decoration on every page). I'm planning ~6 pages: cover → recognizing symptoms (mild vs. emergency, matched to your audience's actual gap) → risk factors specific to landscaping (heat, exertion, PPE, hydration) → a branching scenario where a coworker starts showing symptoms mid-shift and the learner must judge severity and respond → a self-check knowledge quiz → close with an emergency-response quick reference. I could still dig into details like how graphic/severe to make the emergency scenario, or whether you want a downloadable quick-reference card as a takeaway — but I think we have enough to build a strong storyboard. Want me to go ahead, or keep shaping it first?
A: Go ahead and build the storyboard
Mindsmith's Storyboard
Mindsmith generated this storyboard in about one minute:
Objectives
Distinguish early warning signs of heat illness from signs of a heat-related medical emergency.
Identify the occupational and personal risk factors that make landscaping work particularly likely to cause heat illness.
Choose the correct immediate response when they or a coworker shows signs of heat illness, including when to escalate to emergency care.
Opening — full-screen cover
Learner meets: a title card that puts them in the field — hot sun, real stakes, the job they're about to start.
Sets the tone: this is about noticing something's wrong in a teammate (or yourself) before it becomes an emergency.
No overview list or objectives — just the hook. (objective 1, objective 3)
Content 1 — Recognizing the Spectrum: Mild Signs to Medical Emergency
Learner meets:
A short AI-generated video (60–90 sec, no on-camera presenter, narrated animated explainer) walking through how heat illness escalates from early symptoms to life-threatening heat stroke
A banded, high-contrast side-by-side: 'Early Signs' (headache, nausea, weakness, dizziness, thirst) vs. 'Emergency Signs' (confusion, slurred speech, seizure, fainting, hot/dry or heavily sweating skin)
Paired correct-action callouts for each tier (cool/hydrate/rest vs. call 911 immediately and cool the worker)
Learner does: a quick sorting or matching check — sort symptoms into 'early' vs. 'emergency' — to force the discrimination skill immediately after seeing it modeled
Through-line: introduces the idea that heat illness is a spectrum and delay is the enemy. (objective 1)
Content 2 — Why Landscaping Work Raises the Risk
Learner meets:
A visual band on occupational risk factors specific to their job: direct sun, physical exertion, distance from shade/water, PPE and heavy clothing
A second band on personal risk factors: hydration habits, medications, health conditions, not being acclimatized (new to the job = new to the heat)
A short interactive (e.g. accordion or flip-style tile) letting learners explore each factor at their own pace
Learner does: a brief self-reflection/ungraded question — which of these factors apply to you personally this week — to make the risk feel personal, not abstract
Through-line: applies the recognition skill by explaining WHY new hires specifically are high-risk (unacclimatized + outdoor exertion). (objective 2)
Scenario — Your Coworker Starts Struggling Mid-Shift
Learner meets: an Enhanced branching scenario (6–8 nodes, custom generated art per node) set on a real landscaping job — midday, a coworker slows down, looks flushed, says they're 'just tired'
The tempting wrong move: brushing it off, letting the coworker 'push through' or self-manage because admitting a problem feels like slowing the crew down or looking weak — cost: symptoms escalate toward a heat stroke emergency
Learner does: makes sequential judgment calls — notice something's off, assess severity, decide whether to treat on-site or call 911, communicate with a supervisor
Includes a recovery path: a wrong early choice doesn't dead-end the learner, it surfaces worsening symptoms that force a correction
Through-line: this is the centerpiece — puts recognition AND response together under realistic social pressure. (objective 1, objective 3)
Check — Self-Check Knowledge Quiz
Learner meets: a short graded knowledge check (3–5 questions) mixing symptom recognition, risk-factor identification, and correct-response sequencing
Learner does: answers scenario-style questions (not just definitions) — e.g. 'A coworker is sweating heavily and seems confused. What do you do first?'
Through-line: confirms the recognition + response skill actually transferred, not just the vocabulary. (objective 1, objective 2, objective 3)
Close — Buddy System and Quick Reference
Learner meets:
A second short AI-generated video (45–60 sec, no presenter) reinforcing the buddy-check habit: watch each other, speak up early, never wait
A full-width conclusion band with a compact 'if you see this, do this' quick-reference (early signs → self-care steps; emergency signs → call 911 + cool immediately)
Learner does: a short written reflection — one commitment for how they'll check in on a coworker (or themselves) this week
Through-line: closes the loop by turning recognition into an ongoing team habit, not a one-time fact. (objective 3)
The lesson is ready for your review. What do you think?
A few places I'd consider revisiting, specific to what's actually on the page now:
Page 2 (Recognizing the Signs) — the sorting/matching check currently uses the same symptom list shown just above it in the early-vs-emergency bands. If you want it to feel less like an immediate memory check and more like real discrimination practice, I could swap in a few symptoms that weren't explicitly named in the bands (drawn from the OSHA document) so learners have to reason, not just recall.
Page 4 (branching scenario) — right now the wrong path ("brush it off") recovers cleanly once the coworker's symptoms worsen. I could sharpen the social-pressure angle further by adding a line of dialogue where a supervisor or crew lead pushes back on stopping work — that's the real-world friction new hires described struggling with, and it's not yet explicit in the scenario text.
Page 6 (closing) — the written reflection prompt is open-ended right now. If this needs to plug into a compliance or completion record, I could tighten it to a specific, checkable commitment (e.g., naming one buddy-check action) rather than free text.
I can make any of these changes, adjust the tone, restructure pages, swap out the imagery style, or take the lesson in a different direction entirely — just tell me what you'd like, and let me know if anything looks off when you look it over.
The Course
I evaluated the course using 10 different criteria. Remember, this isn't a true experiment with strict controls. I skipped most of the instructional design process, and the prompt was intentionally pretty generic. I didn't even provide learning objectives. I presented a performance problem and was curious to see what Mindsmith would create if I didn't lead it through the design of the content.
I'll highlight a few of my observations below, but overall, the “out-of-the-box” course scored 22/30 using my self-created criteria.
Content Selection & Performance Alignment: The content and interactivity were built around knowing what to do on the job and making safe choices. I would probably delete the videos that didn't add anything necessary to the module. (I highly recommend not having it create videos with your initial prompt. Instead, specifically prompt the videos you actually want. Use those credits wisely, not on generic videos!)
Content Accuracy: It stayed within the information provided in the OSHA guide and didn't hallucinate any new “information.”
Cognitive Load: It's not visually or textually overwhelming. Content is broken into manageable chunks, with interactions that help learners understand and practice the content rather than overshadowing it.
Feedback: The question-and-answer activity provides specific feedback depending on the answer selected. I would tweak some of the feedback slightly for clarity. The feedback in the scenario interaction was less accurate but easy to edit.
Accessibility
Check out the full YouTube video of my accessibility run-through for more details, but I'll highlight some of the main concerns below. (Keep in mind that Mindsmith is aware of these findings and is working through accessibility issues!)
Headings: There are programmatic headings, but it's difficult to tell what heading level you're creating while authoring. Title becomes a programmatic Heading 1, Heading becomes Heading 2, and Subheading becomes Heading 3. I would love to see the authoring experience line up more clearly with the actual heading structure.
Lists: The initial lists created had selectable checkboxes for no purpose, but the checkboxes did function well. I had the ability to choose from different types of lists, but unfortunately, the resulting content was not programmatically identified as a list.
Focus: Visible focus was almost nonexistent on the Submit buttons, and I couldn't find a way to edit the focus indicator or change the color of the buttons. In a drag-and-drop activity, the draggable elements had no visible focus indicator at all, although the screen reader did announce which item was selected. Everything else I tested had good visible and programmatic focus.
Images/Non-Text Content: The illustrations Mindsmith created for the module had excellent alternative text. I rarely see this on a first pass with AI. Unfortunately, the images themselves are exposed as buttons to enter fullscreen, and the alt text isn't read until the image is opened in fullscreen.
Color & Contrast: Overall no issues. The course title is usually placed over an image with a semi-transparent overlay, and Mindsmith struggled with the contrast there because the contrast depends on the image underneath. This was very easy to fix with a quick reprompt to change the color and test myself.
Videos: The videos all had accurate transcripts and captions. Audio description is not currently supported in Mindsmith. In this particular course, the videos didn't add anything necessary to the learning experience, so I would have deleted them before publishing anyway. One of the videos also had some flashing text for over 9 seconds (and on a screen discussing seizures). Again, I highly recommend prompting specific videos you'd like in your course and not letting it create videos from one general course prompt.
Keyboard/Screen Reader: I didn't encounter any keyboard traps. The drag-and-drop instructions weren't available visually, and the instructions provided to VoiceOver were incorrect. I also encountered inconsistent keyboard behavior in some of the question interactions. (See the video for the full demonstration.)
Using Mindsmith in Practice
Based on this first pass, I see Mindsmith as useful for getting to a reviewable draft quickly. I've completed several small projects with Mindsmith, and when given more concrete performance goals, it does a bit better at creating usable content on the first pass. For an experiment based on my own curiosity and a hypothetical scenario, I wasn't going to dedicate the time it would take to complete a full analysis and provide the detailed prompts needed to create a more polished experience on the first pass.
This is one of the limits of giving any tool source material without the surrounding design context. If the prompt and source don't clearly define the performance goal, the tool has to guess. Most tools will respond by making the assessment match the visible content. That's reasonable, but it's not enough to create effective learning content.
That isn't a small thing. Starting from a blank page can slow a project down, especially when the source material is dense or scattered. A generated draft can give a designer, SME, or stakeholder something specific to react to.
I wouldn't use the generated output as the final learning experience without revision. I'd use it as raw material. The value isn't that it replaces instructional design. The value is that it can produce a draft that makes the design conversation more concrete and can be edited into a publishable learning experience once the accessibility barriers I encountered can be addressed.
The Bigger Accessibility Takeaway
I am a firm believer that AI should not be left unattended, and Mindsmith is no exception. That being said, I'm impressed with the initial content output created from one very general prompt. One of my biggest takeaways from this experiment has less to do with Mindsmith specifically and more to do with using AI to create learning content: you still have to know what you're looking for.
A course can look accessible at first glance. It can have headings, alt text, captions, keyboard access, and visible focus. But the existence of those things doesn't necessarily mean they're implemented accessibly. A heading can exist but have the wrong structure. An image can have excellent alt text that isn't actually exposed to a screen reader. An interaction can technically work with a keyboard while still providing incorrect instructions or feedback.
And those aren't necessarily things you're going to notice if you don't test them.
That's why I think accessibility knowledge becomes more important, not less, as these tools get better at generating content. AI can create an impressive amount of a learning experience very quickly, but someone still needs to know how to evaluate what it produced, recognize where something isn't working, and determine whether it can be fixed before that content reaches learners.
Don't forget to check out my YouTube video to see some of this in action!
Comments