Calvin Iwata Memoji, smilingCalvin Iwata Memoji, celebrating with confetti
ProductApplePreview

Apple

Operations: Global Sourcing

An AI assistant designed so its output is trustworthy enough to act on without redoing the work yourself.

Product screens for this study are shared with the full case study.

The problem

An assistant that summarises your work is only useful if you can trust it enough to skip checking it.

The hard part was never generating text. It was that a plausible-sounding wrong answer costs more than no answer, because the reader has to redo the work to find out which one they got.

So the design problem was confidence, not capability.

The rest of this case study is protected

It covers internal work, so the detail and the results are shared on request. If you were sent here as part of an application, the password is in that note.

Or email calvin.lee@berkeley.edu

More work

All work