Aayushman
Gupta
Principal Product Designer, Cequence
Ted-like
AI & UX
Beginner Designers
Mid-level Designers
About The Speaker
Aayushman Gupta is a Lead Product Designer at The Economist with over 14 years of experience shaping digital products across media, research, and intelligence platforms. With a strong foundation in end-to-end design and agile methodologies, he currently leads strategic initiatives at the intersection of user experience and AI-driven innovation. Passionate about scaling design impact in complex B2B environments, Aayushman is focused on exploring new research models where traditional user access is limited — bridging human-centered design with machine intelligence.
About The Talk/Workshop
Designing for Agentic AI - Navigating UX in the Age of Autonomous Systems
AI is evolving beyond simple assistants to autonomous agents that can plan, execute, and adapt—fundamentally changing how we approach UX design. This talk explores the critical shift from designing interfaces to designing relationships between humans and intelligent systems.
Drawing from real-world examples and practical challenges, we’ll examine the four major UX obstacles in agentic systems: opacity (users don’t understand what AI is doing), misalignment (AI goals diverging from user intent), trust gaps (users hesitant to delegate), and autonomy drift (AI acting beyond intended scope).
Through concrete examples—we’ll explore emerging design patterns including transparent planning, goal negotiation, bounded autonomy, and graceful failure modes.
Attendees will leave with actionable frameworks for creating agentic experiences that feel like capable partnerships rather than replacements, ensuring users maintain meaningful control while benefiting from AI’s autonomous capabilities.
Key Takeaways
Design Relationships, Not Just Interfaces
Traditional UX optimized for task completion. Agentic UX optimizes for ongoing human-AI collaboration. Focus on trust-building over time and partnership patterns rather than linear workflows.
Transparency is Your Superpower
Users delegate more to AI that shows its work. Always show the AI's plan before execution, provide confidence levels, enable interruption points, and explain decisions when things go wrong. Add "show reasoning" toggles to AI features.
Design Boundaries, Not Just Features
The more autonomous AI becomes, the more critical it is to define limitations. Design scope-setting controls, clear escalation paths, graceful failure modes, and always provide "I'll take it from here" options.