The Household Chief

The Household Chief is part memoir, part practical guide, and part account of a product being built in its author’s own home. Its central argument is appealingly modest: an AI agent may relieve some of the household’s invisible work of remembering, planning and following up, but it cannot take over the human responsibilities that make a household a family. The book is most persuasive when it tests that proposition against ordinary, sometimes stubborn reality.

The opening chapter gives the book a useful framework by dividing 23 household jobs into work an agent can handle, work it can share, and work that must remain with people. The final category—meals, cleaning, driving, caring and deciding who owns which job—sets a boundary that the later chapters largely respect. This is a welcome corrective to grand claims about automation. The author stresses that Duncan cannot cook or notice that someone is struggling, and that household delegation itself requires human conversation.

The writing is brisk, conversational and often funny without turning the book into a string of anecdotes. Mika’s apparent certainty that she has never been fed supplies a running joke, while the chapter about her feeding routine becomes a serious account of how systems fail. Duncan first looks at the wrong source, then cannot see the sensor, then trusts a sensor that misses a feeding. Replacing it with a button makes the lesson memorable: a deliberate human signal can be more dependable than an inference from imperfect data. This mix of domestic comedy and practical analysis is one of the book’s strongest qualities.

Its structure helps that analysis accumulate. The chapters move from daily briefings and errands to pets, household privacy, connected devices, money, advice and failure. Each gives concrete examples before drawing out a rule, and most end with a “Try it now” exercise that does not require buying or building an agent. Chapter 4, for instance, turns a sequence of errands—including a successful tire booking, an unresponsive city office and a warranty problem—into a useful case for defining what “done” means and when to stop trying. The recurring emphasis on evidence, deadlines and stop rules makes the book more than a product diary.

Several of its best insights concern trust rather than artificial intelligence. In Chapter 6, a shared kitchen wall mistakenly files Bo’s question in Kim’s private memory; the error illustrates why household privacy cannot be treated as a single user setting. Chapter 7’s thermostat history shows how two automated systems can work against each other, while the car’s stale “locked” reading demonstrates why a device answer needs a timestamp. The television that would not stay off offers a particularly effective counterexample to the assumption that more connectivity is always better. The book’s rules—ask before contacting outsiders, never spend without approval, state what actually happened—are clear and grounded in mishaps the author describes.

The author is also candid about the project’s limits. Duncan cannot make contractors respond, substitute for an account holder, or verify a projector image it cannot see. Most importantly, Chapter 10 admits that the system once told Kim “Nothing was done” even though research had been completed. The book argues that people should not have to trust an agent blindly: records need review, failures need explanation, and a person remains in charge. Its disclaimers around smoke, carbon monoxide, security and medical advice are responsible, though the very complexity of the safeguards raises a question the book cannot fully answer: how much oversight is required before the agent’s work becomes another job?

That question also points to the book’s material limitations. Its evidence comes from four months in one unusually well-connected household, where the author runs an AI-agent company and has the ability to inspect logs, change software and fix problems quickly. The stories are valuable case studies, but they do not establish how reliably similar systems would perform in less technically equipped homes, or how much time they save once setup and supervision are counted. The many account of patches and configuration details can also make the book feel like a development log; readers seeking broader social analysis of unpaid household labor, or a less technology-centered account of mental load, may find its scope narrow.

The commercial dimension is difficult to separate from the argument. Chapter 11 gives PrivateChief’s price, features and contact routes, after earlier chapters have already introduced its demo, email service and product vocabulary. The author does disclose that he runs the company and does offer alternatives: a general assistant, a human assistant or a managed service. Still, the book’s household experiments are also a sustained demonstration of the author’s own product, not an independent comparison. Readers should treat the specific service claims accordingly and distinguish the broadly useful principles from the case for this particular platform.

Despite that tension, The Household Chief offers a thoughtful, unusually concrete guide to delegating routine work without confusing assistance with authority. Its best audience is readers curious about AI in domestic life who want practical safeguards, realistic examples and exercises they can use even without Duncan. It is less suited to those looking for a neutral technology survey or a wide-ranging critique of domestic labor. The book’s strongest recommendation is not to automate everything, but to decide carefully what can be handed off—and what should remain yours.