4HWW Was Right — We Just Needed Better Agents
A few weeks ago I found myself with a free Tuesday afternoon and no idea what to do with it.
That is not a humblebrag. It is closer to a confession. For about a year I have been handing the repetitive, schedulable, attention-poor parts of my work to a small set of AI agents — the fortnightly finance review, the sweep through dormant projects, the monthly audit of our data, the research digests I never used to read. The work that used to sit on a list and quietly rot. The agents, unlike me, actually do it. Every fortnight, on the day I told them, whether or not I am in a meeting, on holiday, or simply not in the mood.
What I had not expected was the free Tuesday. I had, without quite noticing, built the thing Tim Ferriss promised nineteen years ago. And when it arrived, I did not want to use it the way he said I would.
📖 What Ferriss actually argued
The 4-Hour Work Week was published in 2007 and sold a great many copies on the strength of its title. The title did the book a disservice, but we will come back to that.
Strip away the bestseller packaging and Ferriss made a serious argument, organised around an acronym: DEAL. Define what you actually want. Eliminate the low-value work that fills a week. Automate what remains. Liberate yourself from a fixed desk. The diagnosis was right, and it has aged well. Most of what fills a working week is not the work — it is the administration of the work. Status updates, calendar tetris, the third follow-up email, the report nobody reads. Ferriss saw clearly that a great deal of it could simply stop.
The step that never quite worked was Automation. In 2007, automation meant other people. You hired a virtual assistant in another time zone to answer your email and book your travel, and you called it a four-hour week. In practice you did not get a four-hour week. You got a second, worse job: managing unreliable strangers across a language barrier and a nine-hour offset, checking their work because the cost of a mistake landed on you. The mechanism was right. The workforce was human, and flaky, and the overhead of delegating to it often exceeded the work you were trying to hand over.
🤖 The automation layer finally arrived
The thing Ferriss was waiting for showed up about eighteen months ago, and it is genuinely good.
I now have a set of agents, each with a narrow remit, that do the work I could never make myself do regularly. One reads my pension and benchmarks the funds every fortnight. One sweeps my dormant projects and asks, for each, whether it is dead or merely neglected. One audits our data for the small rot that accumulates when nobody is looking. They are tireless. They are cheap. They do not get bored, or offended, or distracted by a more interesting problem. They show up on the Tuesday I asked for, every time, and they do not need managing the way the virtual assistant did, because the cost of checking their output is far lower than the cost of doing the work myself.
I wrote about this recently as cadence — the reliable, repeatable rhythm of attention that humans are genuinely bad at sustaining and agents are genuinely good at providing. That piece was about what the agents give you day to day. This one is about what happens to the time they give back.
Because here is the part worth being honest about: for the first time, the Automation step in Ferriss's framework actually holds. The four-hour week is buildable now. Which brings me back to the free Tuesday, and to the thing he got wrong.
⚖️ Right about the mechanism, wrong about the point
The promise of The 4-Hour Work Week was leisure. Mini-retirements. Learning to tango in Buenos Aires while your automated business ran itself. The implicit theory underneath was that work is a cost — a thing to be minimised, escaped, reduced to the smallest viable number of hours so that real life could begin somewhere else.
Here is what I actually did with my free Tuesday. I did not learn to tango. I spent it on a pricing problem for sheepCRM that no agent could have touched — not because the agent is not clever enough, but because the answer lives in fifteen years of conversations with membership organisations that exist nowhere except in my head and the heads of a few colleagues. It was hard. It took the whole afternoon. It was the best work I did that week.
The agents did not give me less work. They gave me my best work back. They took the half of the job that was draining and schedulable, and handed me the half that is irreplaceable — and that half turns out to be the part I actually want to do.
I argued in The Sum of All Tokens that the cost of software is token spend plus human insight, and that the insight is the scarce half. This is the same equation, pointed at a working life instead of a codebase. Agents drive the cost of the cheap half towards nothing. What is left — what becomes more valuable, not less — is the half that only a person can do. The judgment. The taste. The fifteen years of context. The willingness to care about the thing on a Tuesday afternoon.
Ferriss told you to minimise the hours so you could escape the work. The better deal, the one the agents actually offer, is to subtract the drudgery so you can do more of the work that was the point all along.
🚫 Against the hustle
I want to be sharp about this, because the internet's version of this story in 2026 is loud, and it is wrong.
The going narrative is the one-person billion-dollar company. Fire your team, wire up a fleet of agents, sell the set-up to other people for five thousand a month, collect passive income while the machine runs. There are a hundred posts this month on the seven agent businesses you can start by Friday with zero employees. Someone has registered The 4 Hour AI Workweek as a brand. The whole genre reads like Ferriss with the serial numbers filed off and the humanity removed.
It is the 4HWW cargo cult. It copies the most visible part of the original — eliminate the humans — and misses the thing underneath it entirely. It treats people as a cost to be removed and judgment as a commodity to be automated, and it is wrong on both counts.
I run a company with ten people in it. I have not let one of them go because of an agent, and I do not plan to. The agents did not replace my team. They removed some of the work that was forcing my best people to spend their days on their worst work — the reconciliation, the chasing, the manual report. What is left for them is the part that needed a person all along: the customer who is upset and needs someone who understands why, the design call that will still be running in production in five years, the judgment that does not fit in a spreadsheet.
The hustle framing is relentlessly positive about AI and quietly contemptuous of people. Machines win; humans are overhead. I think that is both ugly and inaccurate. The honest version is positive about both at once. Agents are extraordinary at the work that has a clean interface and no soul. People are extraordinary at the work that has neither. Put the two together and you do not get a smaller business run by one exhausted founder chasing passive income. You get a better business, run by people who have been handed back the work worth doing.
🧭 What the four hours were always for
Ferriss called the book The 4-Hour Work Week, and that title is why so many people misread it. It put the emphasis on the number — on working less, as if less work were self-evidently the goal. But the four hours were never the point. The point was always what you did with the rest.
For me the rest is not a beach (well sometime it is a beach or a hammock in the garden...). It is the pricing problem, and the difficult customer conversation, and the architecture decision I will be living with in year fifteen, and the half-hour with our head of software where we skip the preamble because we have built up fifteen years of shared reality and can get straight to the hard part. That is not a cost to be minimised. That is the work. It is the reason I started, and it is the part I would keep even if I could automate it away — which, conveniently, I cannot.
Ferriss was right. We just needed better agents. And we needed a clearer idea of what the freed-up time was actually for — which was never to do less, but to spend more of ourselves on the half of the work that only we can do.
James Webster is the founder of sheepCRM and director of Croftsware. This piece extends Agents Give You Cadence and The Sum of All Tokens.