Forum Discussion
Managing sensitive data in Storyline simulations, including AI-assisted workflows
Our source systems may contain confidential information that can't appear in simulations.
How are others handling this?
- Synthetic/test data before recording?
- Redaction or replacement after capture?
- Edited screenshots instead of recordings?
- AI-assisted redaction, data replacement, or generation of realistic synthetic data?
We're looking for a workflow that balances security, realism, efficiency, and maintainability.
1 Reply
Depends on the simulation.
If the content is more or less a bunch of UI screenshots dressed up with invisible buttons to emulate the environment, either directly editing the image or adding a textbox over it on the slide with a generic alternative. When I worked at Cigna and we needed to show live claim systems (no training environment) we'd intentionally modify the claim numbers such that they would look real to the untrained eye but if you entered them into a claim system, would not only not return an actual claim but, due to what each of the thirteen digits actually represents, just didn't make sense.
I've done (and indeed am making updates on) courses that rely on scripted videos. Those tend to be less simulations though and more "watch this interface be used." They can be made interactive, but the pain-point is still the live customer data, so in those cases I really, really, really go out of my way to avoid scrolling windows or doing a lot of tabbing. I'm good at editing confidential information out of a video using After Effects, but screencasts that scroll a lot can make the redaction incredibly time-consuming.
The third, best (and hardest) way is to build the system in Storyline. I was part of a team that reproduced an in-house CRM in Storyline. There was a top menu bar with links to the other pages in the CRM. When you entered a phone number, an "account" was loaded by some JavaScript, and a bunch of fields in the fake CRM would populate. Scenario 7 always had the same customer and date of birth. All of the customers lived in a JSON array, and since it was basically a giant phone book/database, adding or updating the fake account info was a piece of cake. No screenshots were ever used so everything looked tight and clear, but super crisp screenshots can be used as a base, and then text boxes with variables can be placed on top of those, styled to the same font and size.
I consider those all different training media, and the ideal media is usually determined by the learning objectives. What does success look like for you all?
