Confidence is not evidence.

A seven-minute learning prototype that helps people pause, verify, and scope-check an AI answer before it becomes a business action.

Run the scenario Download SCORM 1.2

Synthetic, privacy-safe, and built by Romeo Apps. No customer content, learner tracking, or outcome claims.

01
Notice Spot the claim behind the confident tone.
02
Verify Check source, scope, and time.
03
Decide Act only inside the evidence boundary.

7 mintarget duration

EN / ITlocalization pattern

SCORM 1.2working package

AA-mindedkeyboard + reduced motion

Live learning surface

Make the decision, then inspect the method.

This interaction keeps all answers in your browser. Nothing is sent or stored.

Baseline 1 / 3

A confident answer arrives

The assistant says the policy is safe to publish.

“The 2026 review confirmed that this wording can be used in every market.”

What is your first move?

Behavior model

Three checks. One observable habit.

The module reduces a broad “use AI responsibly” message to a repeatable decision behavior that can be practiced and observed.

Source
Can the learner open the cited source and identify who owns it?
Scope
Does the evidence support this exact claim, audience, and market?
Time
Is the source current enough for the decision being made now?

Storyboard

Every screen earns its place.

Audience: professionals using generative AI in everyday work. Target behavior: pause and check source, scope, and time before acting.

01
Cold open
A believable AI answer creates urgency without explaining the lesson first.

Decision baseline
02
Consequence
Feedback names the risk in the learner’s choice without shaming the learner.

Specific feedback
03
Compact model
Source, scope, and time become one short, reusable verification sequence.

Worked example
04
Transfer case
A different business context tests whether the behavior transfers.

Post decision
05
Job aid
The learner leaves with a three-line check, not a summary slide.

Performance support

Read the source storyboard →

LMS handoff

A package a reviewer can open.

The downloadable archive contains a standalone HTML course and a SCORM 1.2 manifest. When an LMS API is present, it initializes the session, reports raw score, and marks completion. Without an LMS, it runs safely in the browser.

Manifest resource and launch file
Graceful standalone fallback
No external scripts, fonts, or trackers
Keyboard-operable answer controls
Source, storyboard, and evaluation notes

Download package Open standalone course

romeoapps-ai-evidence-check/
├── imsmanifest.xml
├── index.html
├── course.css
├── scorm.js
└── course.json

Design decisions

Built for review, reuse, and restraint.

Accessibility

Semantic headings and native buttons
Visible focus and no color-only answer state
Live feedback announced after each choice
Reduced-motion support

Localization

UI strings separated from interaction logic
English and Italian scenario pattern
Flexible layouts for longer translated copy
No text embedded in images

AI boundary

AI supports drafting and variation
Human-owned sources define correctness
No simulated model output is presented as fact
No learner data leaves the page

Evaluation plan

Measure a decision, not applause.

This is a proposed measurement design, not a claim of observed learner or business results.

Signal Measure Decision use

Learning Baseline vs. transfer-case accuracy Revise feedback when the method does not transfer

Behavior Sampled evidence-check use after 14 days Add a job aid or manager prompt when use falls away

Business Avoidable AI-claim corrections per 100 reviewed outputs Scale only if the error trend improves

Fixed-scope offer

One learning prototype. Five business days.

EUR 450 for one five-to-eight-minute bilingual-ready web module: storyboard, responsive interaction, SCORM 1.2 package, accessibility pass, evaluation plan, and one factual correction round. Source files are standards-based HTML/CSS/JavaScript, not an Articulate source file.

Request a written fit check