Friday, 5 June 2026

Travel is structurally unique in facing AI pressure from both sides of the P&L simultaneously: on the demand side, AI agents are exploding look-to-book ratios because — unlike human browsers — they do not stop searching, driving up serving costs for airlines and hotels on every non-converting query; on the supply side, operating internal AI models at scale is also rising in cost despite falling per-token prices. The piece was timed to Skift's Data and AI Summit and draws on analysis showing this dual squeeze has no close analogue in other major industries.

Airbnb CEO Brian Chesky is in the early stages of funding an independent AI lab focused on user interaction and design — not foundational models — and will not lead it day-to-day while remaining Airbnb CEO. The move reflects Chesky's longstanding view that existing LLM products are not yet ready for Airbnb's product needs; unlike Booking and Expedia, Airbnb has not integrated booking tools into major AI platforms like ChatGPT.

A practising designer and design system consultant argues that AI's genuine value in the design process lies in the delivery phase — prototyping, code generation, design system migration — but that discovery (problem framing, user insight, strategic direction) remains irreducibly human. The author, currently migrating design systems for LLM legibility and shifting toward code-first prototyping, frames the current moment as genuinely disorienting: practitioners are relearning process fundamentals while simultaneously shipping production work.