In late 2024, Nielsen Norman Group published a finding that would have seemed paradoxical a decade earlier: mobile-first design was degrading the desktop experience. Content dispersed across endless scroll. Oversized touch targets replaced precise click zones. Images optimized for five-inch screens stretched awkwardly across monitors. The desktop web had become a secondary interface, and secondary interfaces follow a recognizable trajectory. They receive just enough investment to function. Never enough to thrive.
The pattern repeats. Admin panels behind consumer applications are routinely described as afterthoughts in design literature, built from component libraries rather than designed for the people who use them eight hours a day. Internal tools at companies with polished consumer products run a generation behind in usability. The interface that generates revenue gets the investment. Everything else gets maintenance mode.
If agents increasingly mediate how people interact with web services, the human-facing web starts to look like desktop after mobile took over.
The precedents all point toward decay. Desktop web didn't get more thoughtful when mobile absorbed the revenue pressure. Design patterns that worked on small screens were applied wholesale to large ones, because adapting costs money and the business case had shifted. Admin panels didn't become elegant when consumer apps became the priority. They became functional enough, and stayed there. You could tell a hopeful story about the human web relaxing into something calmer once freed from conversion optimization. But nobody has ever observed a secondary interface getting more investment after losing its commercial primacy.
There's a possibility, though, that this case breaks the pattern. The human web and the agent protocol are fundamentally different media. Desktop and mobile were both visual interfaces for the same humans, one getting more budget. Admin panels and consumer apps were both screens, one getting more design attention. A machine-readable capability declaration and a human-facing website share almost nothing in their form. The usual trajectory of gradual neglect might not apply. Something stranger might happen instead.
People are already discovering that interacting with a service through an agent is simply a better experience than using the website directly. Better in a specific, almost physical sense: the absence of friction that existed only to work against you. No confirmshaming language. No visual misdirection. No urgency countdown ticking in the corner. A structured interaction where the available actions are declared plainly and executed without theater. Using a service that isn't trying to manipulate you into staying feels like relief.
This is already happening in small ways. People use price comparison engines rather than visiting airline sites directly, because the comparison engine strips away drip pricing and urgency cues. They use browser extensions that auto-apply coupons rather than hunting through checkout flows designed to make discount codes hard to find. Each of these is a case where someone chose a mediated, more structured interaction over the "designed" human experience, because the designed experience was designed against them.
If that pattern scales, the human web can stay exactly as it is and still lose. It only needs to matter less. Whether anyone invests in an interface that fewer people choose to use, or whether the human door quietly becomes the side entrance that nobody maintains, the history of secondary interfaces suggests the answer. Then again, the history of secondary interfaces has never included one that started out as the entire point.