CASE STUDY
From Storyline to AI-Produced HTML5: Rebuilding a Software Simulation
I rebuilt an outdated Articulate Storyline software simulation as an AI-produced HTML5 activity in under an hour, against a 4-hour Storyline production benchmark. Then I turned the workflow into a reusable skill to streamline production on future projects.
Originally built for Wix's Customer Care onboarding. The portfolio sample is a rebranded recreation with placeholder content.

At a Glance
My role
Learning experience designer: script revision, interaction design, AI-assisted production, QA, workflow design
Audience
Wix Customer Care agents who support users with SEO
Course
Introduction to SEO, an internal onboarding course
Activity
A 24-screen guided click-through of the platform's SEO Checklist tool, built from 14 screenshots
Tools
Google Docs (storyboard), Articulate Storyline and Rise, Claude, HTML5/CSS/JavaScript
Timeline
Under a week from script revision to launch. The HTML5 build took under an hour, with build and QA done in 2 days. The skill was shared with the team within a week of launch.
Outcome
Production time down 75%+ (under 1 hour vs. a 4-hour benchmark). The HTML5 version replaced Storyline in the live course. Future updates take one file plus a revised script.
Before and After: The Workflow
The steps stayed almost the same. What changed is who does them: hand-building slides and triggers moved to Claude, and an automated test runs before my review. Updates now start from the saved file instead of re-authoring in Storyline.
The Challenge
The product had changed so much that the activity was teaching a workflow that no longer existed. The original 2024 simulation walked agents through a navigation path, setup questions, and checklist sections that had all since been redesigned. Test sites showed a different entry button than the script expected, and the script diverged from the live product from about slide 10 onward.
The activity is the hands-on core of the course. Agents use it to practice the exact path a user takes, so they can guide users through it on support calls. An inaccurate simulation means agents give inaccurate guidance.
Rebuilding it in Storyline meant recapturing every screenshot and hand-placing every hotspot, trigger, and state, about 4 hours of production per pass. Every future product change would cost the same again.
Design: Rewriting for the Product As it is Now
I rewrote the storyboard against fresh captures of the live product, keeping the proven "show, then let them click" pattern but rebuilding every step around the new flow. I confirmed the storyboard with the SEO subject-matter expert as the source of truth before production.
Area
Before
After
Practice moments
Repeated copy-and-paste for each keyword
One copy-and-paste for the business name and one for the site description, so practice stays focused
Checklist tour
Sent learners to the site editor and back to refresh
Stays in the checklist: status icons, the progress counter, and fixing one real task (adding a title tag)
Ending
A dimmed overlay on the last screenshot
A distinct "Activity complete" screen with Restart, added after QA feedback
Key Interaction Decisions:
- Click on the real target.
Every "click X" step is a pulsing hotspot on the screenshot. A wrong click nudges the learner toward the target instead of failing silently.
 - Show only what the learner needs to see.
Where a screenshot already showed the answer (a filled-in field, a ticked box), I masked it so the learner does the action first. Where I wanted attention without a click, I dimmed the rest of the screen instead of outlining, because outlines read as clickable.
 - Protect privacy.
I replaced a tester's real name on the dashboard with the course persona and blurred account details on the sign-in screen.
Production: I Designed, AI Built
I built the HTML5 version in under an hour, with Claude doing the coding and me directing the design and checking every screen.
I owned
Claude handled
The revised script and every word learners read
Turning the script into a single self-contained HTML5 file
Screen-by-screen interaction choices (hotspot, Continue, copy-and-paste, mask, spotlight)
Placing hotspots from screenshot coordinates
Visual direction: matching the existing course template
Layout, responsive scaling, full screen, Back and Restart
QA: clicking through every path and catching mismatches
Revisions on request
The activity went through one QA round. QA found that the original ending made it easy to restart by accident, so I designed a distinct completion screen and shipped the same day. It replaced the Storyline in the live course, and the subject-matter expert approved it enthusiastically.
The result also behaves better than the Storyline did. It's one file with no publish step, it scales to any width in Rise, and the layout stays the same from screen to screen.
Making it Repeatable: A Reusable Skill
On my own initiative, I turned the workflow into a reusable Claude skill for the LXD team. Anyone can now turn a script and screenshots into a tested, ready-to-publish activity.
The skill builds each activity from a locked template, so every activity matches the course's look. It also tests itself automatically by clicking through every step. Before it builds, it flags problems for a person to decide on, like screenshots that don't match the script or real names showing on screen. I wrote design lessons from testing into it, such as using a spotlight instead of an outline for emphasis, so the next designer starts from what already works.
Results
Measure
Result
Production time
Under 1 hour, vs. a 4-hour Storyline benchmark (75%+ less)
Shipped
Replaced the Storyline in the live course, under a week after script revision
Stakeholder review
Approved by the subject-matter expert after one QA round
Maintenance
Future updates take the saved file, a revised script, and new screenshots. No re-authoring.
Team reach
Workflow shared as a reusable skill, already used by a teammate
What I'd Take Forward
The hard part of a simulation like this was never the clicking. It was deciding what the learner should see, do, and notice on each screen. Handing production to AI moved my time from placing triggers to that design work.