Talk to Me: Designing Human-AI Conversations

Conversational AI is transforming how people interact with technology. From virtual assistants to customer service bots and AI-powered mental health tools, well-designed AI conversations can enhance user experiences, build trust, and drive engagement. But as we move into an era where human-AI interaction is a core part of daily life, how do we ensure these experiences are intuitive, effective, and valuable?

This course, taught at Stanford d.school in 2021, explored how to design human-AI conversations that work—both technically and emotionally. Alongside co-instructors Mariana Lin (former lead writer for Apple’s Siri) and Pablo Paredes (HCI researcher and faculty at Stanford Medicine), I guided students through a hands-on, prototype-driven studio where they built AI-driven interactions focused on stress reduction and well-being.

Building AI That People Want to Engage With

When humans talk to AI, they expect clarity, responsiveness, and a seamless flow. But most AI interactions still feel awkward, inefficient, or overly scripted. This class examined the design challenges that make or break AI adoption, including:

• How to create AI interactions that feel natural and effective—without misleading users about AI’s actual capabilities.
• How to handle errors, misunderstandings, and ambiguity—building AI systems that can recover gracefully when conversations go off track.
• How to make AI more engaging and useful in personal and professional contexts—from guided coaching to productivity assistants.
• How to design AI that supports well-being, rather than contributing to digital fatigue.

From Concept to Execution

Students worked through fast-paced design sprints to prototype AI-driven conversational experiences. They explored key elements of voice, personality, and interaction flow, iterating on AI personas that could guide, assist, and respond to user needs. The class focused on:

• AI personality and tone: Crafting a distinct AI voice that matches the task, whether authoritative, friendly, or playful.
• Conversational UX: Designing AI that anticipates user intent, avoids friction, and provides meaningful responses.
• Multimodal interaction: Exploring how text, voice, and visual elements can work together for a more seamless AI experience.
• Scalability and implementation: Understanding how AI conversation design integrates into real-world products and services.

The Impact

With AI becoming an integral part of digital products, getting the conversation right is critical to adoption, engagement, and long-term success. This class equipped students with the strategic, technical, and creative skills to design AI interactions that aren’t just functional—but truly valuable.

For designers, product teams, and technologists, mastering AI conversation design is an essential skill for the next wave of digital innovation. This class provided a practical, hands-on approach to building AI that people trust, enjoy, and actually use.