I originally designed this project as a text-based simulation but later transitioned it to a voice-based experience to create a more authentic practice environment. Because dog training relies heavily on verbal communication and real-time decision-making, speaking commands aloud felt much closer to the experience of working with a live dog than typing responses into a chat window.

My goal was to create a simulation that felt dynamic and realistic while acknowledging the limitations of a digital environment. Although learners are not interacting with a real dog, they are required to think on their feet, practice newly learned training techniques, and make decisions in the moment without the pressure or risks associated with working with a live animal.

To further enhance the experience, learner feedback and performance scoring are also AI-powered. Rather than receiving static right-or-wrong responses, learners receive feedback tailored to the specific strategies they used during their interaction with Max. This creates a more personalized experience and encourages reflection, refinement, and repeated practice.

AI Tools Used:

• Devlin.ai was used to create the voice-enabled simulation, AI coaching experience, and adaptive feedback system.

By incorporating conversational AI, I was able to create a more immersive practice experience that allows learners to safely build confidence and apply foundational dog training skills before working with a real dog.

Rather than leaving learners to figure things out entirely on their own, I designed the simulation to include an AI-powered training supervisor who provides guidance throughout the experience. Katie, the training supervisor, serves as a coach who helps learners reflect on their decisions while maintaining a supportive and encouraging tone.

Through prompt engineering and iterative testing, I designed Katie's feedback to reinforce effective training strategies while gently discouraging less effective ones. A key design decision was ensuring that Katie never simply tells learners the correct answer. Instead, she provides hints, observations, and encouragement that help learners think through their next steps and arrive at solutions independently.

This coaching approach mirrors the type of support a novice trainer might receive from an experienced supervisor in a real-world training environment. By combining guided practice with learner autonomy, the simulation encourages active problem-solving while still providing the support needed to build confidence and skill.

To make the experience feel more engaging and responsive, I designed multiple visual states for Max that change throughout the simulation based on learner interactions. As learners attempt to train Max, his expressions and behavior shift to reflect what is happening in the conversation, creating the sense that they are interacting with a living, responsive animal rather than a static character.

This design decision was intended to provide learners with immediate visual feedback and reinforce the impact of their choices. Just as a real dog responds to its environment and a trainer's actions, Max's behavior changes based on the learner's approach. These visual cues help communicate whether the learner is gaining Max's attention, losing his focus, or successfully building toward the desired behavior.

How I accomplished this:

• Used Articulate Storyline object states, variables, and triggers to manage Max's visual state changes

• Used Devlin.ai to change the Storyline variables that triggered changes in Max's  visual and behavioral state

• Iteratively refined state-selection logic to ensure Max's behavior remained aligned with the AI coach's narrative feedback

By combining AI-driven behavior logic with Storyline's visual state management, I was able to create a more immersive and believable simulation experience that reinforces learner decisions in real time.

Creating a believable and instructionally effective AI-powered simulation required extensive prompt engineering, testing, and refinement. Rather than treating the AI as a standalone chatbot, I approached the prompting process as a design challenge—carefully shaping how the training supervisor provided guidance, how Max responded to learner actions, and how performance was evaluated throughout the experience.

Throughout development, I defined how the AI coach should communicate, what behaviors should be encouraged or discouraged, how feedback should be delivered, and when the simulation should conclude. A key design goal was ensuring that learners received enough support to remain successful without having the AI simply tell them what to do. This required careful prompt design to balance coaching, learner autonomy, and meaningful practice.

Iteration played a critical role in refining the experience. Through repeated testing, I adjusted how Max's visual states were selected, how coaching feedback aligned with learner actions, how scoring was calculated, and how the simulation progressed toward completion. These refinements helped ensure that the AI's responses remained consistent with both the instructional goals and the visual experience presented to the learner.

This iterative process allowed me to create a more cohesive and believable simulation in which coaching, feedback, visual behavior, and performance evaluation work together to support authentic skill practice.