Building AI Products That Last: User Control, Creative Workflows, and the Evolution of AI Interfaces

Designing AI user interfaces requires balancing non-deterministic outputs with user trust and control. In this episode, we explore how AI product managers should approach building intuitive interfaces that give users agency while leveraging AI's creative potential, particularly in creative and collaborative workflows.

Guest Introduction

Laura Burkhauser is the VP of Product at Descript, an AI video and podcast creation platform. With deep experience building AI-powered creative tools, Laura brings unique insights into designing user interfaces that feel natural and controllable despite AI's inherent unpredictability. Having led the development of Underlord, Descript's AI co-editor, she offers practical guidance on creating "open world" AI experiences that say yes to users while maintaining safety and usability.

Why AI UI Design is Different

Trust Varies by Domain: The cost of AI mistakes in creative work (bad video edit) is much lower than in finance or healthcare, allowing for more experimental and open-ended interfaces

Non-Deterministic Challenge: Unlike traditional software with predictable outputs, AI interfaces must help users navigate unpredictable results while maintaining confidence

Human-AI Tool Parity: AI should have access to the same controls as human users, preventing situations where AI creates something users can't manually fix

Context Switching Costs: AI's processing time creates new UX challenges around waiting states, collaboration patterns, and maintaining user flow

The "Open World, Infinite Lives" Design Philosophy

Open World Principle: AI should be capable of handling any user request rather than limiting interactions to predefined workflows - avoiding "phone tree in a trench coat" experiences

Infinite Lives Safety Net: Robust undo/revert capabilities give users courage to experiment with AI, knowing they can always roll back problematic changes

Scaffolding Over Restrictions: Instead of limiting AI capabilities, provide templates and examples that guide users toward successful interactions while preserving flexibility

Prompt Transparency: Show users the actual prompts being used, allowing them to understand and customize AI behavior rather than hiding the "secret sauce"

Collaborative AI Interface Patterns

Real-Time Collaboration: Allow users to continue working while AI processes requests, similar to multi-user document editing experiences

Stacked Requests: Enable users to queue multiple AI tasks while maintaining the ability to interrupt or modify the sequence

Community Prompt Sharing: Build systems for users to save and share successful prompts, creating collaborative improvement cycles

Contextual Guidance: Use starter templates and examples to stack the deck toward successful outcomes while maintaining open-ended capability

Managing AI Unpredictability in Creative Tools

90% Problem: Design for scenarios where AI gets users 90% of the way to their goal, then provide human tools for the final 10%

Rule-Based vs. Model Improvements: Balance adding restrictive logic against waiting for better underlying models to naturally improve behavior

Quality vs. Freedom Trade-offs: Sometimes accept temporarily imperfect AI behavior to preserve the open-world experience that will benefit from future model improvements

User Education Through Experience: Focus on getting users to their first "wow moment" with AI rather than over-explaining capabilities upfront

The Evolution Beyond Chat Interfaces

Language as Universal Interface: Chat/voice interfaces leverage humanity's most efficient communication method - tens of thousands of years of linguistic evolution

Multimodal Pointing: The future includes combining language with visual pointing, screenshots, and direct manipulation of interface elements

Voice as Natural Progression: As AI understanding improves, voice becomes viable beyond limited command sets - enabling true conversational interfaces

Tactile Integration: Combining traditional UI manipulation with AI understanding creates more intuitive creative workflows

Practical AI Interface Design Strategies

Template-First Onboarding: Instead of blank prompts, provide pre-filled examples that demonstrate best practices while remaining editable

Progressive Complexity: Start users with successful patterns, then gradually expose more advanced capabilities as they build confidence

Visual Feedback Systems: Use interface design to communicate AI's current state, capabilities, and limitations without explicit instruction

Error Recovery Patterns: Design clear paths for users to recover from AI mistakes without losing their broader creative flow

Lessons for AI Product Managers

Domain-Specific Trust Building: Understanding your domain's failure costs helps determine how much AI agency to provide users

Community-Driven Improvement: Build systems that capture and share user-discovered best practices for AI interaction

Infrastructure Investment: Focus on building robust human-AI collaboration tools rather than just improving AI capabilities

User Mental Model Alignment: Design interfaces that match users' existing expectations for creative collaboration rather than forcing entirely new paradigms

This episode provides actionable frameworks for product managers and designers building AI interfaces, emphasizing that success comes from thoughtful user experience design and clear control mechanisms rather than simply hiding AI complexity from users.

Interested in being a guest on Future Proof? Reach out to the show hosts.