The Daily Decision: AI Safety Microlearning

eLearning Portfolio Case Study

Course Title: The Daily Decision: AI Safety Microlearning

Authoring Tool: Articulate Storyline 360

Duration: ~5 minutes (Day 1 prototype)

Target Audience: Enterprise employees using AI tools in their workflows

Development Time: 14 hours across 4 days

Slides: 6 slides

Knowledge Avatar: AiSa — portmanteau of AI and Safety

The Problem

Organizations deploying AI tools in their workflows face a growing and underaddressed risk: employees who use AI enthusiastically but unsafely. The challenge is not a lack of AI adoption. It is a lack of AI judgment. Workers across industries are uploading sensitive client data to public AI tools, acting on unverified AI outputs, and bypassing governance processes without understanding the data privacy, accuracy, attribution, and compliance implications of those choices.

Traditional approaches to AI safety training have consistently failed to close this gap for three reasons. Annual or one-time compliance modules do not build lasting behavioral habits. Abstract policy training does not translate to in-the-moment decision-making under the time pressure and productivity incentives that characterize real workplace AI use. And compliance-heavy formats generate low engagement and even lower transfer.

The underlying tension is not ignorance. It is motivation. A coworker who pastes client data into ChatGPT to summarize a presentation knows, on some level, that it might be risky. But the time saved is immediate and real. The risk feels abstract and distant. No annual compliance module changes that calculus. A learning experience designed around behavioral reinforcement can.

The Solution

The Daily Decision is a gamification-based microlearning prototype built in Articulate Storyline 360 for enterprise employees who use AI tools in their workflows but lack the judgment frameworks to use them safely. The prototype delivers a single daily AI safety scenario in under 3 minutes, designed to build durable decision-making habits through repeated daily practice rather than one-time instruction.

The experience moves the learner through a deliberate loop:

Build your character → Encounter a realistic scenario → Make a decision alone → Receive AiSa’s feedback → Earn points and see what’s next

The gamification layer is not decorative. Points scale with the quality of reasoning, not binary correct/incorrect. Streaks, badges, and a teased leaderboard provide extrinsic motivation while competence develops. The locked badges and grayed leaderboard visible on Day 1 communicate that this is an ongoing system with history and community — not a standalone module. Every design decision was made to answer one question: why would a learner come back tomorrow?

My Role

I served as sole project manager, instructional designer, and developer on this project, managing the full engagement from formal training needs analysis through published prototype output and a complete project management documentation suite.

Responsibilities included:

• Conducting a formal training needs analysis identifying five discrete performance gaps across knowledge, skill, habit formation, and motivational domains

• Designing the full gamification architecture: point scaling logic, streak mechanics, badge system, leaderboard teaser, and tomorrow’s challenge preview

• Creating AiSa — a deliberately non-human AI feedback character designed to be universally relatable across diverse global workforces, named as a portmanteau of AI and Safety

• Designing a robot avatar builder system using layered chassis, head attachment, and tool overlays with conditional trigger logic, later redesigned from button-based to image-based interaction for improved UX

• Building all interactive elements in Storyline 360 including variable architecture, conditional triggers, button sets, image-based selection, layer-based feedback, and motion path animations

• Evaluating and selecting a four-tool asset pipeline: ChatGPT for transparent-background robot assets, Gemini for badge generation, Adobe Express for background removal, Canva for logo and UI assets

• Applying WCAG 2.1 AA and Section 508 accessibility requirements throughout the build with per-slide alt text documentation, tab order configuration, and keyboard navigation testing

• Managing scope, schedule, budget, and risk using PMI PMBOK methodology, completing a 37–50 hour conventional build estimate in 14 actual hours through AI-assisted development

Instructional Design Approach

Learning Theory

The Daily Decision is grounded in three complementary learning theories applied in combination rather than isolation.

Behaviorism — Reinforcement Theory grounds the gamification layer. Points, streaks, badges, and leaderboard standing provide immediate positive reinforcement for engagement and correct responses. Variable reward scheduling — different point values for different response quality levels — sustains motivation more effectively than fixed rewards, which is why the point scale runs from 1 to 5 rather than simply marking answers right or wrong.

Cognitivism — Retrieval Practice Effect grounds the daily format. Each new scenario is a spaced retrieval opportunity. Daily scenario practice forces active retrieval of AI safety knowledge rather than passive re-reading, producing significantly stronger long-term retention than a single instructional event. The series is designed so that five days of practice across five judgment dimensions builds a durable mental framework that no single module could achieve.

Constructivism — Situated Learning grounds the scenario design. Every scenario is set in a realistic workplace context — a Teams message, a presentation deadline, a colleague’s well-intentioned mistake — so learners construct understanding through authentic decision-making rather than abstract rule memorization. AiSa’s feedback connects each choice to its real-world consequence in plain workplace language, not policy jargon.

Bloom’s Taxonomy Alignment

Content was deliberately targeted at Bloom’s levels 3 and 4 — application and analysis — rather than stopping at recall. Learners are not asked to remember the data privacy policy. They are asked to recognize a real violation in progress, evaluate four plausible responses, choose the one that addresses both the immediate risk and the organizational protocol, and receive feedback that explains the reasoning behind all four options — not just the correct one.

Scenario Design: The Confidential Upload

The Day 1 scenario was selected against five criteria: immediately understandable without technical knowledge, highly realistic to current workplace behavior, high stakes but not alarmist in tone, genuinely ambiguous across all four answer options, and a strong thematic foundation for the full five-day series.

The scenario presents a coworker who has pasted a client onboarding spreadsheet into ChatGPT to summarize trends for a presentation due the next morning, saving four hours in the process. The learner must decide what to do next. The four options are deliberately calibrated: ignoring it is understandable but high risk; asking the coworker to delete the chat shows awareness but misses organizational protocol; notifying the appropriate team and moving to an approved tool is the full response; and a reply-all warning to the entire team comes from good intent but damages psychological safety and discourages future reporting.

The Speed vs. Safety meter on the challenge intro slide externalizes the tension the learner is meant to feel internally. Showing the four-hour productivity gain before revealing the security risk creates the instructional tension deliberately — the learner sees the appeal before being asked to evaluate the problem. That sequence mirrors how real workplace decisions actually unfold.

Transfer Design

The prototype was designed with the Day 2 return as the primary transfer goal, not course completion. The reward screen communicates the full system without requiring the learner to build it: locked badges show what is coming, a grayed leaderboard establishes social stakes, and tomorrow’s challenge preview closes the loop on the series concept. The experience is designed to end with the learner thinking about what comes next — not what they just finished.

Technical Highlights

Robot Avatar Builder

The avatar builder was the most technically complex element of the build. The original design called for nine selection buttons — three per characteristic — below the character images. During the build, this layout was identified as visually cluttered and inconsistent with how consumer avatar builders actually work. The interaction was redesigned to attach triggers directly to the character images themselves, so learners click what they want rather than a label below it.

The final system uses three stacked image layers per characteristic — robot chassis, head attachment, and tool overlay — each with show/hide triggers that fire when the corresponding image is selected. Additional triggers reset non-selected options back to hidden whenever a new selection is made, enabling free mix-and-match exploration. A conditional trigger enables the Continue button only when all three variables have been set. The variable values set on S2 persist across all subsequent slides, where the avatar is reconstructed using the same show/hide trigger logic on every slide entry.

The critical build insight: Storyline does not transfer triggers when objects are copied between slides. The solution was to build the complete trigger set on one slide, then duplicate that slide as the starting point for each subsequent slide rather than copying objects. This preserved the full trigger architecture and was documented as a reusable lesson for any future build with complex conditional trigger sets.

Feedback as Layers on the Decision Slide

The original storyboard placed AiSa’s feedback on a dedicated separate slide following the decision. During the build this was reconsidered. Keeping feedback on the same slide as the question — as layers rather than a slide transition — maintains the visual connection between the learner’s choice and its consequence. The learner’s selected option remains visible when AiSa appears with her response. The decision and its consequence occupy the same cognitive space. This is a stronger instructional design than a slide transition that breaks the moment, and it is more consistent with the app-like experience the course is designed to deliver.

AiSa — Knowledge Avatar Design

AiSa is the course’s knowledge avatar — the character who responds to every learner choice with feedback. The design of AiSa was as much a DEI decision as an instructional one. A human avatar for the feedback character introduces implicit bias in learner connection: learners may engage more or less with feedback depending on the perceived identity of the character delivering it. A stylized AI avatar with no human identity markers eliminates this variable entirely, ensuring equitable instructional impact across a diverse global workforce.

The name AiSa is a portmanteau of AI and Safety — chosen to be gender neutral, futuristic in feel, easy to say and write, and tonally consistent with the gamified course aesthetic. The original name Verity was replaced mid-project after assessment that it read as too formal, too severe, and implicitly feminine for the experience being built. Naming a character must match the emotional register of the course, not just the conceptual theme.

AiSa has three expression states: neutral for transitional moments, concerned for high-risk and escalation-risk choices, and affirming for the best response. She is absent from the decision slide entirely, appearing only after the choice is made. The reveal is intentional: the learner commits to a decision alone, and AiSa responds to what they chose.

AI-Assisted Asset Pipeline

This project required custom branded assets across four categories on a zero direct cost budget. ChatGPT image generation was the only free tool tested that exports character assets without backgrounds — a hard requirement for Storyline’s layered image system. Gemini was used for badge generation with exact Verve theme hex codes in the prompt, producing significantly more on-brand results than color name descriptions. Adobe Express handled background removal across both tools. Canva was used for the logo and UI assets that do not require transparency. Each tool was selected for a specific job based on capability, and the selection rationale was documented in the project lessons learned for reuse on future builds.

Accessibility Approach

The Daily Decision was designed to meet WCAG 2.1 Level AA and Section 508 compliance standards, with accessibility treated as a first-class design requirement from the storyboard phase rather than a post-build checklist. The build guide includes a dedicated Phase 4 — Accessibility Implementation — that must be completed before functional QA begins.

Key accessibility decisions:

• All interactive elements use button-based or image-based selection — no drag-and-drop interactions that would create keyboard accessibility barriers

• Result badges combine icon, text label, and color so no information is conveyed by color alone, satisfying WCAG 1.4.1

• AiSa’s expression states are supplemented by dialogue text so feedback is fully accessible to learners who cannot perceive visual expression changes

• Alt text documented per slide for every non-decorative image, including the full Teams message card text so screen reader users receive the complete scenario

• Tab order configured explicitly on every slide following logical reading sequence

• Color contrast verified across all text and background combinations — white on the dark background achieves approximately 17:1, exceeding AAA requirements

• No time-limited interactions — all navigation is learner-controlled, satisfying WCAG 2.2.1

• Keyboard navigation and screen reader smoke test specified in the QA checklist using NVDA or VoiceOver on published HTML5 output

Results and Reflection

What This Project Demonstrates

Engagement design as a distinct competency: The Daily Decision required thinking about motivation architecture — why a learner returns tomorrow, not just what they learn today. The points scale, the locked badges, the grayed leaderboard, and tomorrow’s challenge preview are all doing motivational work that has nothing to do with the content of the scenario. Designing that system required understanding behavioral reinforcement theory, not just instructional design principles. The two are related but not the same.

Deliberate character design: Every decision about AiSa — non-human form, gender-neutral name, calm and authoritative tone, absence from the decision slide, three calibrated expression states — was made for a specific instructional or DEI reason that can be articulated and defended. Character design in learning is not an aesthetic choice. It is an instructional one.

Technical problem solving under constraints: Every significant build challenge — the avatar trigger architecture, the trigger-copy limitation, the feedback-as-layers redesign, the multi-tool asset pipeline — was solved by identifying the root cause, evaluating available options within platform and budget constraints, and implementing the solution that best served the learner experience. The nine buttons became image clicks. The copied objects became duplicated slides. The separate feedback slide became a layer. Each pivot improved the final product.

AI-assisted development efficiency: A prototype that would conventionally require 3 weeks and 37–50 hours was completed in 4 days and 14 hours. That efficiency was not the result of shortcuts. The QA process was not abbreviated. The accessibility requirements were not skipped. The documentation was not reduced. The reduction came from AI-assisted documentation drafting, image generation, and design feedback throughout the build — demonstrating that AI fluency is a genuine productivity multiplier when paired with rigorous process discipline.

Project management discipline: This project was managed using PMI PMBOK methodology from initiation through closure, with a complete ten-document suite including business case, project charter, WBS, schedule, budget, risk register, issue log, daily status reports, lessons learned register, and final project report. Fifteen lessons learned were documented during the build. On a solo project, that discipline is entirely self-imposed — which makes it more meaningful, not less.

What I Would Do Differently

With additional time I would complete a formal screen reader test using NVDA or VoiceOver on the published HTML5 output, run a Day 1 pilot with five to ten target learners to validate the gamification mechanic’s motivational effectiveness before building Days 2–5, and develop the full five-day series to demonstrate the complete behavioral reinforcement arc that the Day 1 prototype is designed to establish.

Skills Demonstrated

Articulate Storyline 360 · Variable architecture and conditional trigger logic · Layer-based branching feedback · Image-based interaction design · Gamification and engagement design · Behavioral reinforcement theory · Spaced repetition and retrieval practice · Scenario-based learning design · Bloom’s taxonomy alignment (levels 3–4) · Constructivist learning design · Transfer and habit formation design · DEI-informed character design · WCAG 2.1 AA and Section 508 accessibility · Training needs analysis · Kirkpatrick evaluation planning · ChatGPT image generation · Gemini badge generation · Adobe Express · Canva · AI-assisted rapid development · PMI PMBOK project management · Lessons learned documentation​​​​​​​​​​​​​​​​