Technology Adoption Is Not Just an Engineering Problem
Every failed technology company built something that worked. Few of them built something people were willing to change their lives around. That gap — between functioning and adopted — is where cultural anthropology lives, and most of the technology industry still treats it as a nice-to-have.
In 1979, a Berkeley anthropology student named Lucy Suchman walked into Xerox's Palo Alto Research Center to study how people used the machines the engineers next door were building. The story that circulated for years was that she found a copier so confusing that someone had to slap a giant green "make it work" button onto it. That version isn't true — the button was already on the machine before she started. What she actually found was harder to put on a slide: people don't follow the plan a system embeds in its interface. They improvise around it, moment to moment, reading cues no designer anticipated. Suchman called it situated action. It became a foundation of human-computer interaction as a field. It's the reason "user experience" exists as a job title at all.
Intel figured out the same thing two decades later, when it made Genevieve Bell — a cultural anthropologist with a Stanford PhD, not an engineer — responsible for understanding how its chips would actually get used. She spent her tenure telling Intel something its roadmaps didn't account for: people don't adopt a technology because of what it can do. They adopt it because of what it lets them keep doing, or become. Microsoft, IBM, and Hewlett-Packard built the same function into their labs around the same period, for the same reason. None of it was philanthropy. Ignoring the terrain was expensive.
Nowhere is the terrain more visible than in the gap between products engineered brilliantly that died and products engineered adequately that spread. M-Pesa is the clean case. Kenya in the mid-2000s had no infrastructure advantage over the dozens of countries where similar mobile-payment products stalled. What it had was a banking sector that had collapsed in the 1990s, leaving people with more trust in cash and community than in institutions. Safaricom, the mobile carrier behind M-Pesa, already held over 80 percent market share and the trust that came with it. Its retail agents were required to paint their storefronts the same green as the brand, so a transaction felt like it was happening inside a relationship the customer already had — not with a stranger. Years after launch, when researchers asked people why they'd adopted it, 78 percent gave a reason rooted in another person: a request, a recommendation, someone showing them how. Among people who still hadn't adopted it, only 18 percent cited a social reason for staying out. It won because it moved through a social structure people already trusted.
This is the part engineering culture resists hardest. A roadmap assumes the world is a set of specifications waiting to be satisfied. Anthropology starts from the opposite premise: the world is a set of practices, obligations, and inherited meanings that a technology has to enter without breaking. Adoption isn't a conversion event. It's an accommodation — the product bending to fit a life that was already fully formed before the product showed up.
AI is about to relearn this the hard way. Every enterprise rollout that measures success in seats provisioned and none in behavior actually changed is repeating the PARC mistake at scale — mistaking the plan embedded in the interface for the situated action of the person forced to use it. The discipline that saw this coming has been sitting in plain sight since 1979. Most of the industry still hasn't hired it.
