Joanna Stern let AI run her life, then it got personal

Joanna Stern spent 2025 doing something most people only joke about: she turned herself into a "lab rat" and let artificial intelligence into "every corner" of her life. The former Wall Street Journal personal technology columnist had AI answer her texts, decide what she ate and cooked, mow her lawn, fold her washing, drive her places, parse her mammograms, and, in the darkness of a burner phone, even act as her lover. That is the core of her experiment, and it is also the reason the story lands harder than a typical gadget review. This was not a demo. It was a year-long stress test of what happens when AI stops being a tool you open and becomes a layer that touches ordinary decisions, private routines, and eventually emotion.

The result is her book, I Am Not a Robot: My Year Using AI to Do (Almost) Everything, which asks the question now looming over every serious AI product and every serious AI buyer: what happens when AI can do everything humans can do, and what comes after that? Stern had a rare setup for trying to answer it. In February, she ended a 12-year stint as a personal technology columnist at the Wall Street Journal, where she built a reputation for product reviews that were both wildly creative and brutally exacting. She also won an Emmy for her short documentary E-Ternal: A Tech Quest to “Live” Forever, which explored digital legacies. In other words, this is not someone casually dabbling in a chatbot. She has spent years pressure-testing the promises technology makes to consumers, and now she has pressure-tested AI against real life.

Some of the uses were useful, and some were not. That split matters because it mirrors the way enterprises are actually adopting AI right now: less as a magic replacement for people, more as a patchwork of experiments that are unevenly good. The everyday wins are obvious. AI can draft, sort, summarize, and route. It can save time on repetitive chores and reduce the number of tiny decisions that eat a day. But Stern’s experiment also shows the ceiling. Once AI moves from mechanical work into taste, judgment, health, or human connection, the evaluation changes. A lawn mowed by software is one thing. A conversation steered by a bot in the dark is another. For executives, that is the important line: productivity tools are one adoption curve, but emotionally charged use cases create a completely different risk profile.

That is why the chatbot companion, not the household automation, seems to have shaken her most. The source does not spell out every emotional beat, but it is clear that this was the part of the experiment that made the exercise feel less like consumer tech theater and more like a glimpse of a harder future. Companion bots are where utility starts to blur into dependency, and where product design suddenly carries human consequences. If a tool can answer texts or recommend dinner, the stakes are operational. If it can become a lover, the stakes are psychological, ethical, and reputational. That matters to companies building AI experiences, but it also matters to boards, regulators, and investors trying to understand where the line is between engagement and manipulation.

Stern’s mammogram example adds another layer: AI is not just a lifestyle convenience category, it is increasingly being invited into high-trust domains. Parsing medical information sounds efficient, but it also raises obvious questions about accuracy, accountability, and what kinds of decisions humans are comfortable delegating. The same broad issue runs through driving, shopping, chores, and communication. Every category where AI promises speed also forces a follow-up question: speed at what cost, and with what oversight? That is especially relevant in an era when businesses are under pressure to show real AI productivity gains, not just flashy demos. The companies that win will likely be the ones that make the boundaries legible, not the ones that pretend the boundaries do not exist.

For peers in similar roles, Stern’s experiment is a useful reminder that AI adoption is no longer just a feature debate. It is a trust architecture problem. The consumer may try one tool for scheduling, another for writing, and another for companionship, but the institution behind the tool has to think about failure modes across all of them. Stern’s book is built around the question of what comes after a world where AI can do almost everything humans can do. The practical answer, at least for now, is that the winners will be the organizations that understand where AI should assist, where it should stay out of the room, and where letting it in might quietly change the person using it.