The UI Complexity Trap
Why adding useful features eventually makes a human-first application harder to use.
Emlyn O'Regan
A successful application often accumulates features. Each is useful to somebody, and each usually needs a way for a human to find and operate it. Over time, settings pages grow, menus become deeper and workflows acquire more steps.
The paradox is that greater capability can make a product less approachable. Even if an individual feature is well designed, the combined interface imposes a cost: navigation, discovery, explanation, visual clutter and the fear of changing the wrong thing.
Traditional solutions help but don't eliminate the problem. Progressive disclosure hides complexity rather than removing the need to organise it. Customisable toolbars shift the work to users. A simplified interface may frustrate those whose essential feature has disappeared.
The complication is amplified by the different 10% problem: users need different subsets of functionality.
Why agents change the trade-off
An operation does not need a menu item simply because it exists. If an agent can reliably inspect and invoke the operation, the human UI can concentrate on actions for which direct control is demonstrably useful.
The underlying capability is still there; it no longer consumes permanent human-interface space.
This doesn't remove all design challenges. Agent tools still need clear semantics, permissions and good error messages. But the complexity is expressed in a form software can navigate rather than one every human must learn.
See The Disappearing Interface and Hybrid Human–Agent Software.