Nobody asked for the sparkle button

Jon Sabutis, a design lead at a financial technology company, argues that AI features built without a user problem drive people away, and offers four questions to ask first.

Jon Sabutis leads design at a financial technology company and uses AI in his own work every day. He argues that many products ship the same AI chat panel, switched on by default and hard to turn off. In his account, companies bought a capability before finding a problem for it, and he proposes four product questions and tactics for roadmap meetings.

• He cites Microsoft's March 2026 decision to cut back Copilot entry points, starting with Notepad, the Snipping Tool, Photos and Widgets.
• His four questions are who asked, what the feature is named after, whether it would ship switched off, and whether it would survive at full AI cost.
• In meetings, he proposes mapping three real tasks over two weeks and requiring a one-page brief for every AI item. He also suggests shipping new AI features switched off as an experiment.
• His examples of good AI features work where the job already happens, such as Gmail's Smart Compose and Photoshop's Remove tool.

A product team can require every proposed AI feature to name the person and task it serves before work starts, and test demand by shipping it switched off. Sabutis offers these as tactics from his own practice, not as results he has measured.

Derived by Working Surface from the article; more in the Playbook. Source line: An agent is a worker. You cannot hire a worker for a job you have not described, and the description is the part nobody wrote.

26 October 2026: teams now judge AI work by tests a person sets, from pass-or-fail checks on what agents build to plain questions about who asked for a feature.