Case Study — Scenario-Based eLearning

AI Feedback Lab

A precision practice tool for managers in multilingual workplaces — built to turn one of the most avoided workplace conversations into a skill that actually transfers.

Articulate Storyline 360 Custom GPT Vyond Action Mapping SBI Model

Experience It →

Role

Instructional Designer (Solo)

SME

Jonathon Bartlett, CHRO

Audience

Managers in Multilingual Workplaces

Deliverable

Scenario-Based eLearning + Custom AI Partner

Framework

ADDIE · Action Mapping · SBI

The Problem

A Gap Training Hasn't Closed

This project began with a pattern Jonathon Bartlett had observed across 24 years as a Chief HR Officer: managers know feedback matters, but in multilingual workplaces, feedback conversations fail at a disproportionate rate.

It's not a knowledge gap. Managers understand what good feedback looks like in the abstract. The gap is in delivery — when a team member speaks English as a second language, managers either avoid the conversation entirely or give feedback that lands as confusing, vague, or culturally dismissive.

The needs analysis sharpened the target: this wasn't a course about difficult conversations in general. It needed to be a focused, repeatable practice tool for one specific skill — delivering culturally aware, structured feedback using the SBI model.

The Solution

Practice That Feels Real

Scenario-Based Learning Three realistic scenarios set in a multilingual workplace, each targeting a different behavioral barrier identified in the action map.
Custom GPT Conversation Partner A Custom GPT built for each scenario — not a branching menu, but an AI that responds to exactly what the learner says, including redirecting vague language and idioms.
SBI Framework Scaffolding A model example, framing slide, priming reflection, and on-demand job aid that build the framework before learners are asked to use it.
Variable, Unpredictable Practice Because learners can't memorize the "right answer," practice stays genuine — the stakes feel real because the response is never identical twice.

My Process

ADDIE, Grounded in Action Mapping

I used the ADDIE model as the project lifecycle framework — Analysis, Design, Development, Implementation, Evaluation — and Cathy Moore's Action Mapping methodology to keep every design decision anchored to observable behavior rather than information delivery. The central design question was never "what do managers need to know?" It was "what do managers need to do differently, and what is getting in the way?"

Needs Analysis

Finding the Real Barrier

Through conversations with Jonathon, I built an action map that translated 24 years of HR experience into a single measurable goal and three high-priority observable actions.

Each action came paired with the specific barrier preventing it — and that barrier drove the design. The table to the right shows how the analysis translated directly into practice activities.

This mapping process is what turned a broad training topic into a focused, behaviorally grounded design — and it prevented the most common ID mistake: building a course that teaches what managers already know instead of practicing what they consistently don't do.

Action map developed in Miro using AI-assisted diagramming, based on SME conversations with Jonathon Bartlett.

Observable Action	Barrier	Practice Activity
Avoid idioms & vague language	Used unconsciously; feels natural	Scenario 1: identify & rewrite
Use SBI model	No framework; feedback is intuition-based	Scenario 2: structure feedback in SBI
Deliver culturally aware feedback	Avoidance; fear of causing offense	Scenario 3: practice with AI partner

Design Decisions

Three Barriers. Three Solutions.

Each barrier from the action map became a design constraint — not just a topic to cover, but a specific condition the learning experience had to create.

Barrier 01

Unconscious Idiom Use

↓

A scenario that requires learners to identify and rewrite idiomatic, vague feedback before they can advance. The friction is instructional — learners have to notice what they don't normally notice.

Barrier 02

No Feedback Framework

↓

A scenario built around the SBI model — with a model example, framing slide, priming reflection, and a reusable job aid available throughout practice. The framework is taught, then immediately practiced.

Barrier 03

Avoidance of Difficult Conversations

↓

A Custom GPT partner that responds to exactly what the learner says — including pushing back on vague language, idioms, and cultural assumptions. Branching scenarios can be solved by pattern. This cannot.

🖥️ Storyline + GPT Architecture Replace with published course screenshot or architecture diagram

Solution Architecture

Two Components, One Experience

The AI Feedback Lab uses a deliberate two-component design. Each tool does what it does best — and neither tries to do the other's job.

Articulate Storyline 360

Provides context, teaches the SBI model, frames each scenario, and guides the learner through the experience. Handles structure, sequencing, and SCORM tracking.

Custom GPT

Serves as the conversation partner. Produces variable, contextually authentic responses — and redirects the learner when their language is vague, idiomatic, or culturally inappropriate.

The GPT is not a feature — it is the practice environment. Storyline sets the scene; the GPT creates the conditions that make deliberate practice possible.

Scenario Design

Scaffolding Decisions

The Scenario 1 storyboard reflects deliberate choices about when to give learners information, when to withhold it, and when to give them agency — because those three things are not interchangeable.

Framing Slide

"Watch Twice. Notice Once." — primes learners to observe rather than evaluate, reducing cognitive load on first exposure and keeping them inside the scene.

Priming Reflection Placement

Moved to after the model example — not before. A choice is only meaningful when the learner has the tools to make a real decision.

Optional Job Aid

The SBI job aid is available as a floating reference throughout practice — not required reading. Scaffolding you don't need should stay out of your way.

Vyond Branching Reserved

Branching video is reserved for Scenarios 2 and 3, where learners have the framework. In Scenario 1, branching would give the appearance of agency without the conditions that make it instructionally meaningful.

Framing Slide

Model Example

Priming Reflection

Practice + GPT

Replace placeholders with published Storyline screenshots

I had the pleasure of beta testing Leslie's AI-enabled performance support tool. With over 24 years of management experience, I found this tool to provide exceptional insight into performance management — and into how rarely we think about language when we think we're thinking about feedback.

Jonathon Bartlett — Chief HR Officer & Community College Professor

Results

Evaluation Plan

As a portfolio project, formal Kirkpatrick data is not yet available. However, the evaluation plan is built into the architecture:

Level 1 (Reaction) — Post-session reflection prompt embedded in Storyline. Level 2 (Learning) — GPT conversation quality assessed through structured debrief questions on the final slide. Level 3 (Behavior) — Manager self-report survey at 30 and 60 days post-training. Level 4 (Results) — Reduction in avoided feedback conversations tracked through HR systems.

What I Learned

Key Takeaways

Building this project taught me that the most important instructional design decisions happen in the analysis phase — not in Storyline. The clarity of the action map made every subsequent choice easier. When you know exactly what behavior you're targeting and exactly what's blocking it, the design almost writes itself.

It also confirmed something I believe about AI in learning: the goal is not to use AI because it's impressive. The goal is to use it only when it does something that no other tool can do as well. A Custom GPT conversation partner earns its place here. It cannot be replicated with branching scenarios.