By playing, you agree to uListen's Terms and Privacy Policy and YouTube's Terms of Service.
EO Studio100 AI Leaders Explain How to Build AI That Will Win in 2026 — WHAT BUILDERS SHOULD DO NOW
Summaries, chapters, transcripts, and FAQs on this page are made by uListen, not YouTube.
At a glance
WHAT IT’S REALLY ABOUT
AI product winners in 2026 will be defined by UX
- The panel argues that “chat-only” interfaces are a UX dead end and that winning AI products will adopt new interaction primitives like structure, semantic resizing, remix, format translation, and better agent monitoring.
- As AI models become commodity infrastructure, competitive advantage shifts to ergonomics-like UX: clarity, safety, trust, and experiences that solve real problems rather than flashy gadgets.
- Builders should treat each LLM as a distinct “material,” starting from what the technology can uniquely do now while iterating back and forth with genuine user needs.
- Case examples (e.g., Grammarly’s evolution from correction to composition, and the Spark character demo) illustrate that emotional resonance, care, and “feeling seen” can drive adoption more than raw capability.
- User research must adapt to AI interfaces by studying prompts and conversations, measuring satisfaction, and designing systems that proactively surface help before users know what to ask.
IDEAS WORTH REMEMBERING
5 ideasStop treating chat as the default end-state UI.
Chat is universal but often linear and limiting; leaders highlight structured experiences (e.g., research reports with sources) and purpose-built interfaces as the next step in AI product design.
Design around emerging primitives: resize, remix, translate formats.
Future products will let users adapt content to context (semantic resize), freely recombine styles and ideas (remix), and move between media (format translation) with minimal fidelity loss.
Agents need “lobbies,” not just background automation.
If AI is running tasks across timescales, users need clear ways to start, monitor, and re-engage—otherwise downtime and ambiguity become the enemy of good work.
Treat each model like a different material with constraints.
Teams should “play with” models to learn strengths/weaknesses and identify the “why now” capability that makes a product newly possible, then iterate with user needs.
Earn the right to go beyond the user’s initial question.
Users ask for the tip of the iceberg; great AI first nails the explicit request, then helps uncover the underlying goal (e.g., venue → full birthday planning) and explores branches like a real thought partner.
WORDS WORTH SAVING
5 quotesChat is both universal and kind of a dead end for user experiences.
— Jess Holbrook
Agents, agents, agents, agents, agents is, is all the stuff right now. It's like all these things existing at wildly different timescales and just multitasking forever at ever-expanding scales.
— Jess Holbrook
A powerful AI model with poor UX is kind of like a heavy duty power drill with a terrible handle.
— Ayça Cakmakli
Anything that violates that, we do not do, and anything that supports that, we say yes to.
— Andy Szybalski
We say ship to learn, right? Uh, but shipping speed may not equal learning speed.
— Jenny Lo
SPEAKERS
Jess Holbrook
cameoAI product/design leader discussing emerging UX “primitives” for generative AI beyond chat.
Ayça Cakmakli
cameoUX lead at Google focused on human–AI interaction and AI product ergonomics.
Pasquale D'Silva
cameoAI product builder/designer speaking about prototyping with LLMs and an AI product called “Cove.”
Andy Szybalski
cameoAnimator/storyteller describing the creation and use of the “Spark” character for memorable AI demos and pitches.
Jenny Lo
cameoProduct strategy and user research practitioner focused on faster learning/iteration in AI products.
High quality AI-generated summary created from speaker-labeled transcript.
