NOTE: Due to confidentiality agreements, the app name and proprietary details have been omitted. This is a real-world case of prompt engineering and conversational UX design for a Spanish-language wellness app.
The goal was to design a voice-based onboarding experience using a virtual guide named Luz. She welcomed users, explained key features, and guided them step-by-step while adapting her tone and language to build trust and reduce user anxiety.
The experience was tailored for Spanish-speaking adults, many of whom were first-time users of AI. Luz used emotionally supportive voice prompts that complemented on-screen text, and celebrated each milestone users completed during onboarding.
I led the concept, UX strategy, and prompt development for this AI-powered onboarding experience. My responsibilities included designing conversational flows, writing adaptive prompts, and creating a flexible UI system that supported multilingual voice and text interactions across mobile platforms.
Goal:
Introduce an AI assistant that feels approachable, helpful, and personalized from the very first interaction.
Behind the Prompts:
The onboarding flow began with a simple yet emotionally intelligent introduction from the AI. I explored multiple tones—from a neutral assistant to a supportive guide—using GPT-based tools to test various conversation styles. After evaluating responses in both Spanish and English, I refined the prompt to include subtle affirmations, reassurances, and a clear explanation of the AI’s role. The result was a tone that builds trust quickly, while staying brief and scalable for different languages.
Initial Greeting (Text + Voice)
“¡Hola! Soy Luz. Estoy aquí para ayudarte a crear tu cuenta y explicarte cómo aprovechar esta app para sentirte mejor.”
Translation: “Hi! I'm Luz. I'm here to help you create your account and show you how to use this app to feel better.”
Purpose: Build trust and set an emotional tone from the first screen.
This is the introduction of Luz. However, the Onboarding has 14 pages. Luz appears only 4 times, but she tells the user what to expect on each page.
Goal:
Help users who feel stuck or hesitant during onboarding with intelligent, well-timed AI guidance.
Behind the Prompts:
Instead of overwhelming users with praise, I crafted lightweight motivational prompts tailored to specific milestones. I tested language that felt natural and supportive across multiple languages and adjusted for tone and cultural nuance. Phrases like “You’re doing great—just one more step!” and “Almost there. You’ve got this.” were designed to celebrate small wins. These prompts help users feel seen and acknowledged, while encouraging autonomy. They reinforce a user-driven experience, offering the flexibility to skip or pause steps without pressure.
Emotional Companion Selection
“Hola, soy Elisa. Seré tu acompañante emocional. Puedes cambiar mi voz en cualquier momento desde tu perfil.”
Translation: “Hi, I'm Elisa. I'll be your emotional companion. You can change my voice anytime from your profile.”
Purpose: Personalize the experience and give users emotional control.
This screen allows the user select an AI agent.
This screen allows the user change Elisa voice tone. Elisa's avatar actually blink and smile to the user.
Goal:
Encourage users to continue by celebrating small wins with motivational, personalized messages.
Behind the Prompts:
Rather than overwhelming users with praise, I designed prompts that offer lightweight motivation tailored to specific milestones. These were tested to sound natural and sincere in different languages and tones. By chaining AI responses and evaluating which variations made users feel more seen and acknowledged, I selected affirmations like:
Onboarding Completion
“¡Perfecto! Tu acceso ha sido confirmado. Gracias por confiar en mí. Estoy feliz de acompañarte en este camino.”
Translation: “Perfect! Your access has been confirmed. Thank you for trusting me. I'm happy to join you on this journey.”
Purpose: Emotionally reinforce completion and encourage continued use.
Final screen of the Onboarding process
To support the prompt-driven experience, I developed a modular, responsive design system tailored for AI-first interactions. The components—dialog cards, input fields, and voice/text toggles—were optimized for clarity, accessibility, and cross-language support.
The system ensures consistency in tone and interaction patterns across mobile devices, making it easier to adapt to both voice and text-based user journeys.
Designing prompt-based AI experiences challenged me to think beyond interface patterns and focus on micro-interactions that build trust, adapt to user behavior, and respect emotional context. I learned to prototype conversations just as carefully as visual flows, ensuring multilingual support and tone consistency across voice and text. This process reinforced the value of designing for flexibility, making the AI feel human, without overwhelming users or disrupting their momentum.
Tested with: GPT-4o (May 2025)
