AI UX beyond chat
IMO AI UX is not chat. I like chat for quick ideas, but for teams it collapses structure, loses memory, and lives outside the work. Better metaphors are agents that act, pipelines that run in the background, and native controls inside the tools we already use. š¤š
Why now: In 2024 and 2025, the wins I keep seeing ship as embedded features, not chatbots. They live where the work already happens. Pull requests. Channels. OS text fields. The browser. They add structure, persistence, and policy. Less prompt, more product.
My pick: @GitHub Copilot Autofix inside pull requests. Security findings turn into suggested code changes with explanations. You accept or edit like any review comment. No chat, just workflow. GitHub says fixes land about 3x faster and up to 7 to 12x on XSS or SQLi.
Other strong non chat patterns:
⢠@Slack AI recaps and thread summaries surface decisions and owners right in the channel. Structured context you can cite, though super basic.
⢠@Apple systemwide Writing Tools let you rewrite, proof, and summarize inside Mail, Notes, and third party apps. OS native control, not a bot window. Still no improved Siri.
⢠Gmail Help me write and @Superhuman AI sits next to Compose to draft and refine replies using thread context. Still AI slapped on top of a rich text editor.
⢠Arc Search Browse for Me runs as a background pipeline that reads the web and returns a structured page. @PerplexityComet actually doing thins for you is possibly one of my favorites.
Design notes for teams that want to go beyond chat:
⢠Keep people in flow. Put AI where the action is.
⢠Prefer explicit controls over long prompts.
⢠Make state visible. Inputs, outputs, and cost.
⢠Log everything for audit. Tie behavior to policy and permissions.
⢠Support handoff between silent automation and user review.
As of August 2025, I don’t think we have figured out the UX for AI.
All I can say is that it’s not chat.