How I Built This: Putting an AI Coach Inside Articulate Rise

What I Built and Why It Matters

I was tagged in an eLearning Heroes discussion asking if anyone had figured out how to build an AI-powered role-play experience directly inside an Articulate Rise course. The post caught my attention because the requirements sounded very familiar. The experience couldn't send learners to another website, require another login, or take them outside the Rise frame.

I've spent a lot of my career designing learning for an enterprise environment, and I've dealt with this kind of problem before. At my previous company, we worked with multiple external vendor partners, each with its own IT restrictions. If something took learners outside our LMS, the site often had to be whitelisted. Getting one vendor to approve a site didn't mean another vendor would. So I understood the problem. I also thought it would be a fun challenge to see if I could make it work.

I'd recently been experimenting with AI in Storyline, so I wanted to see if I could do something similar in Rise. And that's where the project started.

👉 Try it yourself: check out the interaction here.

Scoping the Concept

The original question was about an AI-powered role-play, but I didn't think I needed to build a full conversational experience to prove the concept. Instead, I focused on an AI-powered coaching activity based on CPNI (Customer Proprietary Network Information). This gave me an opportunity to use AI for something that a traditional knowledge check doesn't handle particularly well.

Instead of asking learners to choose the right answer, I wanted them to write what they would actually say to an employee they were coaching. The AI could then evaluate that response against a consistent set of criteria. 

I kept it to one scenario on purpose. This was a proof of concept, not an attempt to build an entire course. If I could demonstrate that a learner could review a scenario, write a response, receive an AI-generated evaluation, and do all of that without leaving Rise, that fit the bill.

The learner takes on the role of a compliance supervisor, reviews a call-monitoring summary describing a CPNI verification failure, and writes the coaching documentation they would give the employee. An employee handbook is available as a reference. Once they submit their response, the AI returns a score from 1 to 10 along with feedback explaining what the learner did well and what could be improved.

I built the course frame in Rise first. Then I used Claude to help plan the interaction. Once we had a plan, I moved into Cursor and provided the prompts, data, and commands needed to build and deploy the application.

I'd used Cursor, Claude, and Vercel for other projects, so none of the tools were completely new to me. There was still a learning curve, though.

Here's the basic setup:

While the technical setup may sound complicated, Claude and Cursor do the heavy lifting, and the experience for the learner is seamless. From their perspective, it's simply another activity inside the course.

Keeping it inside the Rise Frame

The biggest question for me was whether I could actually keep the interaction inside the Rise frame. My first attempt was to simply use the deployed URL in Rise's Multimedia Embed block but during testing, Rise opened the application in a new browser tab.

💡 The fix: Adding an iframe code around the URL kept the application inside Rise.

That was the moment I knew I had the proof of concept I was looking for. No separate login. No external site for the learner to navigate. Just the activity sitting inside the course.

Designing the AI evaluation

Next was identifying what I wanted AI to evaluate and what useful feedback should look like. After identifying what a good response should contain, defining the evaluation criteria, and determining how the feedback should be presented, I used Claude to help develop and refine the scoring prompt and then used Cursor to build the application around it. I tested the interaction, looked at the results, and adjusted things when they weren't behaving the way I expected.

👉 For the technical side: I documented the process in a beginner-friendly build guide. It walks through the tools, setup, API connection, scoring prompt, deployment, and Rise integration.

Key Decisions and Trade-offs

A few decisions shaped the final experience.

One scenario, not a whole course

I wanted to prove the concept without turning it into a giant project. One scenario was enough to demonstrate the interaction. In a real learning experience, I could see this becoming a series of short practice activities rather than one large course.

Open-ended response, not a knowledge check

This was probably the biggest instructional decision. Storyline and Rise can already handle multiple-choice knowledge checks. There wasn't much value in adding AI simply to determine whether someone selected the right answer. An open-ended response is different. The learner has to formulate an answer, and AI can evaluate that response against defined criteria. That's where I felt AI actually added something.

Typing, not voice

I considered allowing learners to speak their responses because speaking can produce a more natural, stream-of-consciousness answer than typing. Ultimately, I decided typing was the better choice for this proof of concept. Voice input introduces additional browser and device considerations, especially when you're trying to create something that works across different environments. Most mobile devices already have speech-to-text built into the keyboard, so I didn't see a reason to add another layer of technology just to solve that problem.

Stay inside Rise 

Keeping the experience inside Rise was probably the least negotiable decision. The original problem was about keeping the learner in Rise. I didn't want to solve the problem by creating a workaround that technically worked but disrupted the experience. I've seen plenty of technology that worked technically but felt disconnected instructionally. I wanted this to feel like part of the course.

Resources and Assets

Here are the pieces of the project if you'd like to explore it yourself.

Key Takeaways

For me, there were three big takeaways from this project.

1. Start small. You don't need to build an entire AI-powered course to see whether an idea is worth pursuing. A small proof of concept can answer a lot of questions.
   
   
2. Use AI where it adds something. I wouldn't use AI for a basic multiple-choice knowledge check just because I could. In this case, the ability to evaluate an open-ended response made the technology useful.
   
   
3. The technology shouldn't drive the learning experience: I started with the learner experience I wanted and then figured out how to make the technology support it.

Where I'd Take It Next

If I were to take this beyond the proof of concept, I'd explore:

• Multiple scenarios with branching: A learner could respond to one situation, receive feedback, and then move into another scenario based on how they performed.
  
  
• Microlearning and practice: I could see this working well as a quick practice activity before an actual coaching conversation.
  
  
• xAPI tracking: I'd like to explore xAPI or something similar to see whether the results could be captured and used beyond the individual interaction.
  
  
• Voice input: Not because I think every interaction needs it, but because it could be interesting to compare how learners respond when they type versus when they speak.
  
  
• Other types of practice: The underlying approach could also be adapted to other types of practice: customer conversations, sales objections, difficult manager conversations, compliance situations, or technical troubleshooting. The scenario and evaluation criteria would change. The basic idea could stay the same.

Have Questions?

I'd love to hear what you're doing with AI in Rise.

• Have you tried embedding a custom AI interaction into a Rise course?
• Have you found another way to keep learners from leaving the course?
• Do you see a use for this kind of open-ended AI evaluation that I haven't considered? 

Share it in the comments below! Questions or comments, look me up on LinkedIn. I'm happy to talk through what I learned building it.

 

Editors Note: Generating a Scenario in Articulate 360 AI

Since this was built, Articulate released native AI Scenario generation in Rise, branching conversations with AI dialogue, and automatic scoring, with no custom code required. It's a great option for preset-choice practice. 

 

🌟 Want to share your build?

Do you have a project you’d love to share with the community?

We’re always looking for more How I Built This stories. Whether it’s a game, interaction, or unique design, we’d love to feature your process. Drop a note in the comments or reach out to the community team if you’re interested!