Skip to main content

Building an Escape Room with AI as Your Instructional Design Partner: Part 2

Lessons learned from using AI as a collaborative partner in immersive instructional design.

In Part 1 of this series, I shared the pedagogical principles that made my virtual escape room successful: narrative structure, consistency, progression, and clear learning goals. Those principles are essential—but they only tell half the story.

The other half? How I actually built the entire project in days rather than weeks.

When I decided to redesign one of the most boring weeks in my course, I didn't have time to start from scratch. I also didn't have a background in game design or creative writing. What I did have was a willingness to experiment with AI tools as collaborative partners in the instructional design process.

This article documents my journey using multiple AI tools to transform traditional course content into an immersive virtual escape room—including what worked, what failed spectacularly, and the workflow I developed for effective human-AI collaboration in course design. If you read Part 1 and thought "this sounds great, but I don't have that kind of time," this article will show you how AI can help.

The Starting Point: Feeding AI Your Content

I began with the raw materials every instructor has: learning outcomes, slide decks, and a general sense of what students needed to learn. My content focused on four academic perspectives (Artist, Humanist, Social Scientist, Natural Scientist) and how they relate to our general education program.

My first attempt: I dumped everything into Claude, learning outcomes, slideshow content, course objectives, and prompted: "Turn this content into an escape room activity."

The result: Four generic activities with vague connections. When I tried to refine with follow-up prompts, Claude kept regenerating similar unhelpful ideas.

Lesson 1: General-Purpose AI Tools Need Specific Context

The problem wasn't Claude's capability, it was my approach. I had given it content without constraints, structure, or examples. Generic AI tools work best when you:

  • Provide examples of what you want (or don't want)
  • Give it a specific role ("You are an instructional designer specializing in game-based learning")
  • Break large requests into smaller, focused tasks
  • Iterate rather than expecting perfection on the first try

But here's what I learned: Sometimes switching tools matters more than perfecting your prompt.

Finding the Right AI Tool for the Job

After hitting a wall with Claude, I discovered the Assignment Idea Bot, a custom GPT created by my colleague Kristen Chorba specifically for generating course activities. This tool was trained on instructional design principles and educational activity frameworks.

The difference was immediate. The same content that produced generic results in Claude generated specific, pedagogically sound activity ideas in the Assignment Bot. The suggestions had clearer learning connections, more detailed mechanics, and actually felt usable.

Lesson 2: Specialized AI Tools Outperform General Ones for Specific Tasks

This was a turning point in my understanding of AI-assisted course design. General-purpose chatbots can do many things adequately, but specialized tools trained on domain-specific knowledge produce dramatically better results.

Consider creating or finding specialized tools for:

  • Assignment generation (custom GPTs, education-specific bots)
  • Rubric creation
  • Quiz question writing
  • Accessibility checking
  • Learning objective alignment

The time invested in finding or creating the right tool pays dividends in output quality.

Multi-Tool AI Workflow: Layering Capabilities

Once I had better activity ideas from the Assignment Bot, I discovered ThingLink's built-in AI tool called Scenario Builder. I fed it my content and it generated:

  • A narrative framework for my escape room
  • Image descriptions for each "room" students would explore
  • Transition language between activities
  • Suggestions for puzzle mechanics

I then took these narrative suggestions back to the Assignment Bot and asked it to strengthen the connections between rooms, ensuring activities built on each other rather than feeling random.

Lesson 3: The Best Results Come from Multi-Tool Collaboration

Think of different AI tools as specialists on a design team. In my workflow:

  • Assignment Bot: Generated pedagogically sound activity structures
  • ThingLink's Scenario Builder: Created narrative cohesion and visual planning
  • Assignment Bot (second round): Refined activities based on narrative framework

Each tool contributed its strength, and I served as the project manager synthesizing their outputs into a coherent whole.

Practical workflow tip: Keep a document tracking what each AI tool generated and how you modified it. This creates a reusable template for future projects.

Advanced AI Application: Creating the Final Puzzle

With the basic structure complete, I wanted a culminating challenge. I returned to the Assignment Bot with a specific request:

"Design a final puzzle that brings together four code words from previous rooms using a cipher code that students can decode."

The bot generated a cipher and suggested breaking it into four fragments to hide throughout the earlier rooms. Good start, but not quite right.

Second prompt: "Create a cohesive final message that ties the learning together, then translate it using the cipher code."

The result: Random numbers that didn't match the cipher at all.

Third prompt: "The numbers don't match your cipher. Can you verify and correct the translation?"

The result: A working cipher that correctly decoded to my intended message.

Lesson 4: AI Makes Mistakes, Verify Everything Technical

This experience taught me a critical lesson about AI-assisted design: Always double-check outputs that require precision. AI tools are excellent at generating ideas and frameworks, but they can produce confidently incorrect technical outputs.

Verification checklist for AI-generated elements:

  • Mathematical calculations (scores, points, cipher codes)
  • URLs and hyperlinks
  • Citation accuracy
  • Code functionality (if using AI to generate interactive elements)
  • Accessibility features (alt text, color contrast ratios)

When AI gets something wrong, simply pointing out the error often leads to a corrected version, but you have to catch the error first.

The Complete AI-Assisted Design Workflow

After building this escape room, here's the workflow I now use for AI-assisted course design projects:

Phase 1: Content Analysis and Ideation

  • Gather all existing materials (learning outcomes, lectures, readings)
  • Use a specialized educational AI tool to brainstorm activity formats
  • Generate 3-5 different approaches before committing to one
  • Use prompts like: "What are five different ways to teach [concept] that emphasize [learning outcome]?"

Phase 2: Narrative and Structure Development

  • Use AI to help create narrative frameworks (if relevant to your activity)
  • Generate transition language and connection points
  • Ask for structural outlines: "Create a step-by-step progression for students"
  • Refine with follow-ups: "Make the transitions more explicit" or "Add a hook at the beginning"

Phase 3: Asset and Content Creation

  • Use AI image generators or AI-powered search tools to find visual assets
  • Generate puzzle elements, questions, or challenge content
  • Create supporting materials (instructions, hint systems, rubrics)
  • Prompt example: "Write student-facing instructions for this activity at a [grade level] reading level"

Phase 4: Technical Implementation

  • Use AI to troubleshoot platform-specific issues
  • Generate code snippets if needed (for H5P, HTML, etc.)
  • Create assessment rubrics aligned to learning outcomes
  • Important: Test everything AI generates before deploying to students

Phase 5: Iteration and Refinement

  • Use AI to generate student perspectives: "What might confuse students about this instruction?"
  • Ask for accessibility improvements: "How can I make this activity more accessible?"
  • Refine pacing and difficulty: "Is this too challenging for first-year students?"

Tools and Platforms: What I Actually Used

AI Tools:

  • Claude: Initial brainstorming (with mixed results)
  • Assignment Idea Bot (Custom GPT): Primary activity design and pedagogical structuring
  • ThingLink Scenario Builder: Narrative development and visual planning

Implementation Platforms:

  • ThingLink: 360-degree interactive space creation
  • Adobe Stock: Sourcing 360-degree images (using AI-generated descriptions as search terms)
  • Adobe Express and Firefly: Creating visual puzzle elements

Why this combination worked:

  • Multiple AI tools provided different perspectives and strengths
  • Implementation platforms had built-in AI features that integrated smoothly
  • I maintained creative control while leveraging AI for acceleration

What AI Can and Cannot Do

After this project, here's my honest assessment of AI's role in assignment design:

AI excels at:

  • Generating multiple options quickly (brainstorming partner)
  • Creating narrative frameworks and story elements
  • Producing first drafts of instructions, rubrics, and descriptions
  • Suggesting structures and progressions
  • Identifying potential accessibility issues when prompted
  • Troubleshooting technical problems when given specific error information

AI struggles with:

  • Understanding your specific students and institutional context
  • Making pedagogical judgments about difficulty and appropriateness
  • Technical precision (codes, calculations, links)
  • Originality that doesn't feel derivative
  • Knowing when it's wrong

Humans are essential for:

  • Defining learning outcomes and pedagogical goals
  • Making design decisions based on student needs
  • Verifying technical accuracy
  • Adding contextual relevance and authentic examples
  • Quality control and final refinement
  • Ethical oversight and accessibility verification

Getting Started: Your First AI-Assisted Design Project

If you're ready to experiment with AI in your course design, here's my advice:

Start small. Don't redesign your entire course. Pick one assignment, one activity, or one week that needs improvement.

Find specialized tools. Search for education-specific AI tools rather than starting with general chatbots. Look for custom GPTs, education-focused platforms, or AI features built into tools you already use.

Prompt iteratively. Your first prompt won't produce perfect results. Plan for 3-5 rounds of refinement, getting more specific each time.
Document your workflow. Save your prompts, track which tools worked best, and note what needed human correction. This becomes your template for future projects.

Verify everything. Especially technical elements, citations, and anything students will interact with directly.

Maintain pedagogical control. AI is a tool that accelerates your instructional design process—it shouldn't drive your educational philosophy or replace your expertise. Remember the principles from Part 1: narrative, consistency, progression, and clear learning goals should guide your decisions, not AI suggestions.

Bringing It All Together

In Part 1, I shared the pedagogical principles that made my escape room engaging for students. In this article, I've shown you the AI-assisted workflow that made it feasible to build in days rather than weeks.

The key insight? AI didn't make instructional design easier, it made it faster. I still needed to:

  • Define clear learning outcomes (Part 1)
  • Make pedagogical decisions about structure and difficulty (Part 1)
  • Ensure narrative coherence (Part 1)
  • Verify technical accuracy (Part 2)
  • Test the student experience (Both parts)

What AI gave me was velocity and variety. I could explore five different approaches in an hour instead of days. I could generate narrative frameworks without staring at a blank page. I could troubleshoot technical issues with a knowledgeable partner rather than googling alone.

The future of course design isn't AI replacing instructors, it's instructors leveraging AI to spend more time on the creative, pedagogical, and human-centered aspects of teaching while offloading the mechanical and time-consuming tasks to tools that excel at those functions.

If you're curious about AI-assisted course design, start experimenting now. The tools will only improve, and the skills you develop collaborating with AI will become increasingly valuable in education. Just remember: you're the instructor or instructional designer. AI is your very capable, occasionally confused, but remarkably productive assistant.

Your Next Steps

Whether you're redesigning one week or rethinking an entire course, the combination of sound pedagogical principles (Part 1) and efficient AI-assisted workflows (Part 2) can help you create more engaging learning experiences without burning out.

If you're ready to build your own escape room:

  • Review the design principles from Part 1
  • Choose one week or unit that needs revitalization
  • Experiment with the AI workflow outlined in this article
  • Remember: you're the instructional designer; AI is your assistant

The tools will only improve, and the skills you develop collaborating with AI will become increasingly valuable in education. Start small, iterate often, and maintain pedagogical control. Your students will benefit from the engaging experiences you create, and you'll discover new ways to bring your content to life.

Tools Referenced in This Case Study

AI tools: Claude, custom GPT assignment bots (for brainstorming and puzzle creation), Adobe Firefly
Platform: ThingLink (for 360-degree interactive spaces)
Assets: Adobe Stock (360 images), Adobe Express (puzzle graphics)