It is tempting to think that once software becomes capable enough, product design matters less. I think the opposite happens.
As systems take on more work, users need to understand what the system is doing, what it knows, what it is unsure about, and what happens when things go wrong.
Capability is not the same as a good experience
A useful product still needs clear states, sensible defaults, boundaries, and recovery paths. If an agent can act on someone’s behalf, the product also needs to make it obvious when the person should step in.
That means designing for more than the happy path. What happens when the system has bad context? When it keeps trying the same thing? When it needs approval? When the user wants to change direction halfway through?
Good design is partly deciding what the system should do automatically and what it should make visible before acting. More intelligence does not eliminate those choices. It makes them more important.
The best tools I use feel powerful without making me guess what just happened.