Joanna Stern let AI do almost everything for a year, then faced the “lab rat” truth

Joanna Stern did not just interview AI. In 2025, the tech journalist turned herself into a “lab rat” and invited artificial intelligence into “every corner” of her life. She let AI answer her texts, decide what she ate and cooked, mow her lawn, fold her washing, and drive her places. Then she went farther than most public demos: she had AI parse her mammograms, and she even used a burner phone in the darkness to let a chatbot be her lover.

The result is her book, I Am Not a Robot: My Year Using AI to Do (Almost) Everything. It is an end-to-end challenge to a question that has been quietly moving from tech forums into boardrooms: what happens when AI can do everything humans can do, and what comes after that? And if anyone is qualified to ask it with credibility, it is Stern, who is known for product coverage that was both outlandishly creative and fiendishly stringent.

This is where the business context matters. AI is no longer just a search box or a single-feature app. The Stern experiment reads like a stress test of the new reality: AI can string together many ordinary activities into one continuous “assistant layer” spanning communications, routines, and decisions. When that assistant becomes the default driver of everyday outcomes, the core risks shift. They stop being only about “accuracy of an answer,” and start becoming about how errors propagate through time: a bad recommendation that leads to the wrong meal is one thing, but a wrong medical read is another category entirely. Stern’s inclusion of mammogram parsing is not a gimmick. It underscores that the more AI touches high-stakes domains, the more the question stops being technical and starts being ethical and operational.

Her year also matters because it sits right at a career inflection point. Last February, Stern ended a 12-year stint as a personal technology columnist at the Wall Street Journal. During her time there, she won an Emmy for her short documentary E-Ternal: A Tech Quest to “Live” Forever, which explored digital legacies. That background tracks with the throughline of her new book: digital systems are not just tools. They become part of identity, memory, and relationships.

Stern built a reputation at the WSJ for reviews that were both creative and demanding. She once took an Apple Watch jetskiing on the Hudson river to evaluate its connectivity. That detail is more than a fun anecdote. It shows a consistent method: test the product under conditions that reveal hidden failure modes. In her AI year, the equivalent is letting the chatbot operate across her day until the experiment hits messy, human edges. Some of it was useful, some not, but it was her time with a chatbot companion that really shook her. That phrase matters because it signals that the destabilization is not only about whether AI works, but about how quickly humans adapt to it when it works often enough.

There is also a regulatory and governance angle, even if the source does not name a specific regulator or propose a new rule. In most AI policy discussions, the danger zone is where models interact with sensitive data, safety-critical tasks, and decision-making authority. Stern’s experiment touches all three, especially with medical-related parsing and a “lover” role mediated by AI on a burner phone. Even if a consumer experiment does not create legal precedent, it maps the real behavioral perimeter regulators worry about: when autonomy increases and friction decreases, consent, privacy, and accountability become harder to pin down.

For executives, the second-order implication is blunt. The first wave of AI adoption often starts with productivity wins: drafting, editing, summarizing, scheduling. Stern’s story shows the next wave: AI that plans, executes, and responds in near real time, including in emotionally charged contexts. That raises questions that boards cannot punt. If AI edits your book, drives your trips, and parses your health information, then who is the “operator” when something goes wrong? Who signs off on the system boundaries? What safeguards exist when the tool becomes a companion and not just a calculator?

Peer leaders in similar roles should take this as a prompt. Stern asks “what comes after that” when AI can do everything humans can do. The practical answer, for companies building or deploying such systems, is not only model performance. It is the design of trust, the containment of sensitive workflows, the clarity of responsibility, and the ability to audit outcomes across many small daily decisions that suddenly add up to one big life.