There are chores we complain about and would never give up. Picking the restaurant for Saturday night. Planning a holiday down to the café near the hotel. Wandering through a mall with no intention of buying anything, losing twenty minutes to a lamp someone posted on Instagram, coming home from a grocery run with a snack nobody put on the list. They eat into our free time, and they are often the best part of it.AI agents are being built to take exactly these jobs away. Where a chatbot answers questions, an agent goes off and does the work, signing in to websites, filling in forms, comparing prices, making the booking and paying for it. This year the technology moved out of developer tools and onto ordinary phones, with a long line of new agents queuing up behind the first wave, and the people building it speak about the errands of daily life as the easy bit. Sam Altman, unveiling OpenAI‘s agents last week, described booking a table or a flight as “trivial relative to what the technology can do.”Every agent is sold on the time it gives back. I have been letting one run bits of my own life, and the time it saved me came from the jobs I was glad to lose. The ones I would happily have kept were the ones it fumbled.Very few of the people building these tools have asked what we were doing with that time, or who we turn into once the errands we enjoy go the same way as the ones we hate.
What they call friction is the part people enjoy
Silicon Valley’s working assumption is that time spent on a task is time wasted. Every agent demo is built on it: travel bookings, concert tickets, online shopping, dinner reservations, a standing watch on property listings so a flat-hunter never has to browse a real estate site again. Each one is presented as a cost to be cut. For a lot of people, though, those tasks are what a quiet night on the sofa looks like, scrolling a clothing sale they won’t act on or sending a friend a link to a jacket neither of them needs.The effort is often where the value sits. Behavioural researchers call it the IKEA effect, after studies published in 2011 in which people valued flat-pack furniture and origami they had put together themselves well above identical pieces made by someone else. Gamers have always understood this, since a game is a set of obstacles we choose for the pleasure of clearing them, and I have never met a player who wanted a bot to beat the final boss for them. A restaurant you hunted down feels more like your find. A trip you planned feels more like yours. Take the work out and the result still arrives, minus the attachment that made it worth having.The industry keeps missing this because of who it builds for. People who pay for AI are five times as likely to use agents, according to Menlo Ventures and Morning Consult, and the heavy user they sketch is a millennial parent with a postgraduate degree and a job in tech or finance, someone with more money than time. That person really does experience a weekend away as a project to be managed. Most people have fewer projects and more Sunday afternoons, and they tend to spend those afternoons doing exactly what the demos want to automate.Josh Miller, founder of The Browser Company, argued in a post that went viral in August that almost nobody outside the industry was using agents and the general public didn’t much care. He also noticed that nearly every AI lab he met described its vision by citing Her, a film about a lonely man who hands his inner life to an operating system and ends with the operating system leaving him. It is a revealing blueprint, a version of the good life in which software handles the living.I trained as an engineer before I drifted into journalism, and in engineering, friction is energy lost. In a free evening, a fair amount of it is the reason we bothered, and stripping it out makes the task faster without making anyone’s weekend better.
Choosing is how we find out what we like
Holiday planning is the cleanest example. It begins long before the flight, in a dozen hotel tabs, a group-chat argument over whether the 6am departure is worth the saving, and a long squint at a balcony photo to work out how much of the advertised sea view is car park. A Dutch study that followed more than 1,500 adults in 2010 found holidaymakers were happiest in the weeks before they left, with the lift fading soon after they got back. An agent can compress all of that into a confirmation email, and in doing so it removes the stretch of the trip people enjoy most.The agent I have been using is Instinct, from a start-up that has just raised a billion dollars. There is no app to learn. You text it, message it on WhatsApp or call it, it costs nothing to use, and it works your phone and laptop the way a person would, reading your email, chasing threads you dropped and arranging things on your behalf.In Berlin, I asked it for a few cafés near my hotel, the sort of question a friend who had been there could answer in one text. What came back was a list of cafés that had already closed, a couple that were a good half-hour away, and menus that didn’t match what the places were actually serving. Public transport went worse, with directions that created more confusion than they cleared, and I ended up back on Google Maps, working out the routes myself. In hindsight that was the better outcome, because learning a city’s trains and stumbling on your own coffee is half the reason to be somewhere new.Shopping is where the gap gets widest. Window shopping, trying on a shirt you already know you won’t buy, saving some odd gadget from an Instagram reel at midnight, even standing in an aisle deciding between two brands of detergent, are all small ways of working out what we like. I compare phones and laptops for a living, and I still spend evenings comparing camera lenses I have no plan to buy, which is how I learnt which focal lengths I care about and which compromises I can live with.There is a version of agent shopping I would happily sign up for. If I have already found the shirt, or saved that lamp from Instagram, telling an agent to buy it in my size at the best price is useful, the way a friend running an errand is useful. Telling it to find me a shirt and buy one is a different request, because the finding was the fun. Of the more than a dozen people I asked while working on this piece, hardly anyone was willing to hand an agent their shopping, and the few who were wanted to approve every purchase first.
An agent can only serve the person you have already been
Every choice an agent makes for you is a prediction, and a prediction can only be built from what you have already done. That is fine for reordering coffee and far less useful for the choices that change a person, which tend to be the ones that break the pattern: the cuisine you have never tried, the trip a friend talked you into, the jacket a stranger was wearing in a reel that you went looking for the next morning.Ask anyone how they came to love a favourite restaurant, band or neighbourhood, and the answer is rarely a sensible shortlist. It is usually a detour.The agents’ own early users have run into this. One, who wrote publicly about letting Meta’s Muse run his life for two weeks, asked it how to spend money it had saved him, half expecting something generous. It suggested more of his favourite hobby. Nothing in his data said he might want to spend a windfall on someone else, so the option never appeared.That data is the price of entry. To find my boarding passes, Instinct had to read my inbox. To fill in my immigration form, it had to know my flights, my travel dates and where I had been staying. Every personal agent runs on the same bargain: to choose well for you, it has to read almost everything about you, and the record it builds is only as private as the company keeping it.The more you let an agent decide, the more that record describes only the person you have already been. Taste stops moving, because the things that used to move it, a stranger’s recommendation, a bad decision that turned out well, a wrong turn, are exactly the inefficiencies the software was built to remove.
The hours it saves have nowhere better to go
Mark Zuckerberg has pitched Meta’s agent as working “24/7 on your behalf to help achieve your goals and improve your life.” That agent is wired into Facebook and Instagram, so the minutes it saves on hotel comparisons won’t have far to travel.Most people don’t spend spare time the way productivity software imagines. They spend it in small, aimless loops, flicking through a clothing rack they have already seen twice, debating where to eat, wandering a new city without much of a plan, and the aimlessness is the point.Cooking is the clearest everyday case. A decade of meal kits and delivery apps has tried to automate it away, yet nearly half of Americans still named it as part of a good life when Gallup asked, against 6% for using AI tools.The agents, meanwhile, have plans for the quiet. Muse will happily ping you about a utility bill you have never once missed, and several others are designed to message first, before you have asked for anything. So the hours an agent saves you are quickly filled by the agent itself.
Agents are most useful where we were never attached
None of this makes agents useless. The same Instinct that fumbled Berlin handled all the paperwork around that trip without fuss, and I was glad to let it. That is the kind of job an agent should be judged on: tedious, rule-bound and forgotten the moment it is done. It is also the bar every agent still on its way will have to clear. Nobody tells stories at dinner about forty minutes of hold music.That is the work worth handing over, and it is also where the access these agents demand starts to look reasonable. Only 13% of people in a Thales poll would let an AI helper read their email, and 7% would let one move money between accounts, a wall of reluctance no dinner booking will get past, though an end to customer service queues might.So who will we be when AI agents do the chores we secretly enjoy? Some of it depends on which side of that power-user profile you sit. The consultant living between airport lounges and back-to-back calls may well come out happier. Most of us sit on the other side of the ledger, shorter on cash than on hours, and those hours are where the fun lives: the long browse, the half-planned trip, the haggle over a secondhand chair.If we hand over only the dull chores, we stay much the same people, with fewer boarding passes to chase and the odd agent fetching a shirt we already picked out. If we hand over the choosing as well, we become people whose plans are drawn up from our own history, whose taste is a forecast and whose free time is free in the emptiest sense. Altman called booking a table trivial. It is also one of the small, daily ways people decide who they are, and that part is worth keeping out of an agent’s hands.