In a column in January concerning the paradox of labor, I recalled the immortal Douglas Adams joke about working circumstances: the hours are good, however “a lot of the precise minutes are fairly awful”. The joke is again already — and generative AI has flipped the script.
Lecturers at UC Berkeley’s Haas College of Enterprise have been doing ethnographic analysis into how expertise staff are utilizing generative AI. Some will let you know that ethnographic enterprise analysis is each the worst sort of enterprise analysis and the worst sort of ethnography, however I admit to a comfortable spot for these items. What the researchers discovered was the other of Adams’ morose Vogon guard: the minutes are wonderful however the hours are horrible.
“In micro moments of prompting, iterating and experimenting, individuals talked about momentum and a way of expanded functionality,” researcher Xingqi Maggie Ye defined. “However once they stepped again and mirrored on their broader work expertise, a distinct tone typically emerged. They described feeling busier, extra stretched, or much less in a position to totally disconnect.”
These tech staff felt that generative AI was making them dramatically extra productive and succesful — however they have been additionally attempting to do extra, voluntarily working longer hours, and hurtling in direction of burnout.
Are these ethnographic observations a glimpse of the long run for the remainder of us?
Little question we will discover out, however whereas we wait, each financial concept and the historical past of expertise have some issues to show us.
Concept first. Think about a contract programmer, paid by outcomes, who used to work 10 hours a day and all of the sudden finds that they will obtain the identical leads to two. Frequent sense would possibly recommend that the coder will begin to benefit from the pleasures of a two-hour workday, however financial concept is extra ambiguous: the “revenue impact” means that the employee ought to work fewer hours as a result of they will obtain a lot by working so little. The “substitution impact” says that staff ought to work longer hours, as every further hour yields bountiful rewards.
Then there may be the query of what the brand new equilibrium can be as soon as everybody masters the expertise. As an analogy, think about that a number of alchemists uncover tips on how to flip lead into gold, however their methodology is quickly being copied. They need to make and promote as a lot gold as attainable, as quick as attainable, earlier than the collapse within the gold market. Coders armed with good AI brokers could also be in the identical place: code as a lot as you’ll be able to when you can nonetheless cost cash to take action, as a result of code could quickly develop into as low-cost as mud.
There’s additionally a company dynamic to think about. It could be that 9 out of 10 in-house programmers are about to be sacked, leaving a handful in cost to handle the coding brokers. If that’s the case, the crucial is evident: to maintain your job, display which you could out-code everybody else within the constructing. Winner-take-all dynamics usually are not a recipe for lengthy lunch breaks and lengthy weekends.
That’s the idea, however historical past has a number of classes for us, too. Visible aids have been as soon as produced by graphic designers and used on particular events solely; the invention of PowerPoint meant that extremely paid and expert professionals began losing time making their very own slides, badly. E-mail is vastly faster and cheaper than a letter, however that merely means a profusion of low-quality, low-value messages bleeding into the evenings and weekends. The library photocopier allowed a era of scholars to repeat educational articles at a velocity their dad and mom might hardly have imagined — nevertheless it didn’t make studying, pondering or studying any quicker.
In every case there was an astonishing enhance in a slim measure of productiveness, however the general impact was to distract from the actual activity at hand, to create a bloated pile of busywork, and to accentuate the sense of productiveness debt, with the checklist of duties individuals felt responsible about not doing getting longer, not shorter.
What the UC Berkeley ethnographers discovered is unusually acquainted. “Employees more and more stepped into obligations that beforehand belonged to others,” they wrote. That’s the unhealthy slide drawback another time.
“As a result of AI made starting a activity really easy . . . staff slipped small quantities of labor into moments that had beforehand been breaks.” All people who lived by the rise of smartphones will nod in recognition.
“Extra multitasking. AI launched a brand new rhythm during which staff managed a number of energetic threads without delay . . . This created cognitive load and a way of all the time juggling.” Nicely, sure — what number of browser tabs do you’ve got open proper now?
I don’t imply to recommend that AI is ineffective or trivial, however there’s a lengthy historical past of time-saving digital applied sciences that at finest make us extra productive but overwhelmed — and at worst, simply make us really feel overwhelmed.
Digital instruments don’t should work this manner. The Nobel laureate economist Claudia Goldin coined the phrase “grasping jobs” to explain roles corresponding to these in company regulation or funding banking the place disproportionate rewards are paid to these keen and in a position to work lengthy hours and be on name at any time when required. She contrasts these with well-paid positions in pharmacy, paediatrics, main care and veterinary medication, the place the roles and the IT methods that assist them have been designed to permit extremely certified practitioners to work restricted hours after which hand over to an equally certified colleague.
It’s not unimaginable to think about AI brokers getting used to facilitate this handover course of, however the discourse in the meanwhile is of good, idiosyncratic human conductors overseeing a frenetic orchestra of AI brokers. Handover protocols sound much less enjoyable however could also be much more helpful.
And what are the remainder of us to do whereas we look forward to the wizards of Silicon Valley to stoop to constructing such prosaic instruments? Todd Brown, a efficiency advisor and managing accomplice at Subsequent Motion Associates, has lengthy espoused preserving an “Agenda” checklist for colleagues and necessary purchasers — working by the checklist face-to-face somewhat than firing off emails at any time when one thing pops up. Now he does the identical for ChatGPT, “with concepts for prompts”.
It could sound like an odd observe, on condition that generative AI — not like a colleague — is all the time accessible. However it is smart. Simply because you’ll be able to flip to AI at a second’s discover doesn’t imply it’s best to. There’s something to be stated for planning forward earlier than interacting with the AI, and for blocking out time with out it — leaving house for the human within the loop to cease, to mirror and to breathe.
Written for and first revealed within the Monetary Instances on 4 March 2026.
I’m operating the London Marathon in April in assist of an excellent trigger. In the event you felt in a position to contribute one thing, I’d be extraordinarily grateful.
