Workers Spend As Much Time 'Botsitting' AI As Producing Useful Work, Survey Finds (yahoo.com) 49
"As the use of artificial intelligence spreads across companies worldwide, it is relieving workers of tedious old chores but creating new ones," reports the Los Angeles Times.
"Most people don't realize the amount of time that they're spending working on the tools to get the time savings that they're professing," said Paul Leonardi, Duca Family professor of technology management at UC Santa Barbara." Leonardi is one of the co-authors of the new study published by the Work AI Institute, whose contributors include academics from Stanford University and UC Berkeley. The institute is sponsored by AI company Glean... The research surveyed 6,000 digital workers across the United States, the United Kingdom, and Australia between December and January. The report found that we are in a phase of significant personal productivity gains, but few companies are translating these gains into revenue and business growth. While 75% of individuals reported a boost in productivity, only 13% of the organizations say they have seen significant business gains as a result of AI adoption, the survey found...
The reason the boost in productivity sometimes leads to waste, Leonardi said, is the time people spend correcting the bot's work and gathering the right files, documentation, and tacit knowledge required for it to produce high-quality output. "It's pretty striking the amount of time and effort people are spending," Leonardi said. Most employees now spend over six hours a week of their workday babysitting their work chatbots, the survey said. There is a "thick, mostly invisible layer of human labor holding the whole thing together," the report said. The survey found that for every hour a worker spends getting useful output from AI, they spend roughly another hour making it usable. Of the total time workers spend interacting with AI each week, 37% goes to botsitting, 36% to actually using the tool to produce work.
Part of the reason so much time disappears into botsitting is how often the tools fall short: Workers report that more than a third of AI sessions fail outright, requiring a full restart or substantial rework. Paradoxically, as more workers hand over bigger parts of their jobs to AI, they are offloading personal judgment and responsibilities to the bots. The survey found 41% of workers say they sometimes deliver AI-generated work they couldn't explain if asked... "I think what's happening with a lot of these Gen AI tools right now is we're essentially expecting individual contributors to act as managers," Leonardi said. "They're just managing these AI tools, AI agents, and we're expecting that they'll be able to produce way more, but we're not taking into account all of the work that actually goes into managing."
This problem isn't likely to go away.
"Most people don't realize the amount of time that they're spending working on the tools to get the time savings that they're professing," said Paul Leonardi, Duca Family professor of technology management at UC Santa Barbara." Leonardi is one of the co-authors of the new study published by the Work AI Institute, whose contributors include academics from Stanford University and UC Berkeley. The institute is sponsored by AI company Glean... The research surveyed 6,000 digital workers across the United States, the United Kingdom, and Australia between December and January. The report found that we are in a phase of significant personal productivity gains, but few companies are translating these gains into revenue and business growth. While 75% of individuals reported a boost in productivity, only 13% of the organizations say they have seen significant business gains as a result of AI adoption, the survey found...
The reason the boost in productivity sometimes leads to waste, Leonardi said, is the time people spend correcting the bot's work and gathering the right files, documentation, and tacit knowledge required for it to produce high-quality output. "It's pretty striking the amount of time and effort people are spending," Leonardi said. Most employees now spend over six hours a week of their workday babysitting their work chatbots, the survey said. There is a "thick, mostly invisible layer of human labor holding the whole thing together," the report said. The survey found that for every hour a worker spends getting useful output from AI, they spend roughly another hour making it usable. Of the total time workers spend interacting with AI each week, 37% goes to botsitting, 36% to actually using the tool to produce work.
Part of the reason so much time disappears into botsitting is how often the tools fall short: Workers report that more than a third of AI sessions fail outright, requiring a full restart or substantial rework. Paradoxically, as more workers hand over bigger parts of their jobs to AI, they are offloading personal judgment and responsibilities to the bots. The survey found 41% of workers say they sometimes deliver AI-generated work they couldn't explain if asked... "I think what's happening with a lot of these Gen AI tools right now is we're essentially expecting individual contributors to act as managers," Leonardi said. "They're just managing these AI tools, AI agents, and we're expecting that they'll be able to produce way more, but we're not taking into account all of the work that actually goes into managing."
This problem isn't likely to go away.
Not bot sitting (Score:5, Interesting)
Re:Not bot sitting (Score:5, Interesting)
That would be true if the technology followed the fabled narrative that the epstein island class want it to.
It actually doesn't work like that though, and no amount of Peter Thiel building a device to justify his childish paranoia will change the fact that inference is not deduction. The false positives will become too hard to hide soon... and when that happens the narrative is going to be pretty hard to control.
It's entirely possible that the class of people they are trying to distract and keep in denial is themselves. They don't build the tech, and they don't run it. They just have stories about how it works.
I'm compiling. (Score:4, Funny)
instead of explaining to my boss why he caught me spinning on my chair and stacking office supplies into Jenga towers as "I'm waiting for the compiler to finish" I can now tell him I'm waiting for claude to get back to me.
Starting with the assumption that AI is faster (Score:5, Insightful)
and then I always realize afterwards that I did spent a lot of time making it actually work.
Not to talk about the nasty bugs and the lack of error handling.
Re:Starting with the assumption that AI is faster (Score:5, Interesting)
Exactly. Executives are *so* sure that AI is 5x to 10x faster, that any measurements to the contrary are disbelieved.
After all, these executives have *all* seen how well Claude can spit out a PowerPoint that looks great, they think it *must* be just as good at coding! Never mind that those PowerPoints they just generated, are usually not effective at communicating their points because they have so much fluff that doesn't matter. And never mind that if the PowerPoint is wrong in some small way, it doesn't actually matter, while with code, you can have disastrous consequences for a small error.
Re: Can do something fun too! (Score:2)
Re: (Score:2)
The problems are different. The question is if you rather type a page of boiler plate in half an hour, or watch the LLM doing it in 30 seconds, then type an explanation what you really did expect (people are bad at explaining and LLM are good at doing what they were told and nothing more as they are programmed to follow prompts as good as possible and not to question them), and let it do it again. For this 30 minute boilerplate task you may need the first time you work this task with this LLM 3 iterations,
"tedious old chores" (Score:5, Interesting)
> it is relieving workers of tedious old chores but creating new ones
If it were relieving workers of tedious old chores, it'd probably be more popular.
From what I can see it's doing the fun parts and leaving the shit parts - us checking it did it correctly - to us.
I went into programming because I enjoyed programming. I would imagine that's true of 99% of programmers. You know what's boring? Checking the code afterwards.
Maybe if the genAI companies found ways to use their technology to automate actual chores, like washing up, cleaning the house, or even (not always!) doing the cooking when we come home exhausted, and driving when our idiot bosses force us to do work at an office, instead of programming, making "art", and stealing shit and rewriting it 100 different ways, it'd be more popular and actually a net positive for the world. People might even spend money on it!
If genAI is truly as intelligent as its addicts claim, that ought to be easy, right?
Re: "tedious old chores" (Score:3)
There is also a gap between when you have a pretty clear idea of what change you want to make and actually getting it done with passing tests. AI also fills that gap.
In neither case do you need to check from scratch. You already know what you wanted. Did it do it or not? That's easy to verify.
Yes, there is some fun lost in not working out all the details yourself. But it makes up for it in the
Re: (Score:3)
... In neither case do you need to check from scratch. You already know what you wanted. Did it do it or not? That's easy to verify. ...
Really? Sounds like your experience with coding is very different to mine. I verify my code before deploying it, yet there have been occasions when bugs are reported, because of cases I failed to think of and verify. Sorry, but I do not think code verification is easy.
Re: "tedious old chores" (Score:3)
Re: (Score:2)
Re: (Score:3, Insightful)
> In neither case do you need to check from scratch. You already know what you wanted. Did it do it or not? That's easy to verify.
Just to be clear, you're suggesting the way to test a program is to just run it once and see if it did what you wanted?
Have you ever done software development? Because that's not how it works. In the real world we do unit testing, code reviews, and other things to try to catch every bug. And, believe me, it's a hell of a lot easier to do those things if you already understand
Re: "tedious old chores" (Score:2)
I don't think the people pooh-poohing AI have much experience with it. It fixes bugs and adds small feature pretty darn well. But some oversight is necessary to get great results. We code review our peers' work before it gets merged, and we need to do the same what Claude produces. Some of the code is produces is sloppy. Some is not. Fix the sloppy stuff, or direct your code assistant to fix it. Invest some time learning to get good results from AI code assistants and you will be amazed at how your producti
Re: (Score:2)
Several ways to look at this (Score:1)
1. This extra time is just that we are new to this. As we get more experience and better at it (in conjunction with the tools improving) the time taken will decrease.
2. This extra time is just par for the course (as the last sentence in the summary seems to hint at, though no explanation as to why which it would have been good to see) and once increases in price for tokens over time are factored in it will cost the same amount (or more) for the same output (however now you are also seeing the hidden cost i
Definitely #2 (Score:5, Interesting)
To date, there's no AI success story, outside of pick and shovel vendors. No startup has leveraged AI to disrupt an existing market and become a household name. Netflix famously leveraged the internet to disrupt Blockbuster's stranglehold on home movies....first with DVD by mail and then with streaming. Salesforce, love them or hate them, disrupted many established players.
If AI ACTUALLY improved productivity, smaller companies would come out of nowhere and eat the lunch of more established players by out-innovating them. Some obvious examples are entertainment. Some game studio from some surprising location would come out with AMAZING AAA games at twice the speed and half the cost. Various business platforms would take on the many fat targets: Microsoft, SAP, Oracle, etc...leveraging AI to out-innovate larger competitors.
You and I may suck at AI and improve with experience...but someone out there is waaaay ahead of us....waaaay more gifted and would theoretically be leveraging AI to build massive projects with tiny teams. But for now, the only people making money are selling tools or computer chips or building data centers for this circular AI economic bubble.
Re: (Score:1)
smaller companies would come out of nowhere and eat the lunch of more established players by out-innovating them.
That is actually what is happening right now. You are just to blind to see it.
Some game studio from some surprising location would come out with AMAZING AAA games at twice the speed and half the cost.
To write a game you need an idea first. And: marketing you still need.
But for now, the only people making money are selling tools or computer chips or building data centers for this circular AI econ
Re: (Score:1)
I do not have an idea for a game in my head.
But if you have one, we could team up.
I am pretty sure I make you a nice first try on the game, in no time - without AI if you like, or with.
Unfortunately: with AI development would likely not be much faster. Perhaps the initial eye catcher would be faster.
Re: (Score:1)
Re: (Score:1)
Why should I give you examples?
Why not just join LinkedIn or something and start a job query?
It is not hard.
Re: Definitely #2 (Score:2)
Well wait a minute, like there was a dot com bubble, but not an internet bubble.
AI related valuations are very fucky and speculative right now. That part can't last forever. But yah, the costs will come down over time and it absolutely 100% is not going anywhere and will be in everything. It doesn't mean markets are rational about what it actually means today, because there are a lot of people throwing a lot of money hard in any direction they think has the AIs.
Re: (Score:1)
The point is that "Machine Learning" left the universities.
The mechanisms how to do it, are now well known - not everything mathematically understood, but well known to train your own models.
You hardly find a job offer on LinkedIn that does not require AI skills one way or the other.
Name a household name!!!! (Score:2)
smaller companies would come out of nowhere and eat the lunch of more established players by out-innovating them. That is actually what is happening right now. You are just to blind to see it.
Citation? Who is a household name leveraging AI to upend an established market...who isn't merely just reselling AI, like Claude/Cursor/etc. Has anyone disrupted a non-pure-technology business? entertainment? logistics? retail? transportation? dating?
Who is the grindr/tinder/uber/salesforce/netflix/amazon of the AI age?
All I know of are pick and shovel vendors. People selling AI to you so you can figure out how to make money with them.
Re: (Score:1)
You can google that.
Seriously?
Why ask me who "inserted" AI into dating for example? There are "new" dating platforms that support AI matching.
Funny, that you are smart enough to come to the idea, but to stupid to google it.
Who is a household name leveraging AI to upend an established market
Every company. Are you stupid or what? They are automating their processes with AI.
No idea what you mean with "household", kitchenware? How the funk should I know anything about kitchenware/households?
Do you know Zeiss? T
Kewl insults, but you fail to make your case (Score:2)
You can google that.
Seriously?
Why ask me who "inserted" AI into dating for example? There are "new" dating platforms that support AI matching. Funny, that you are smart enough to come to the idea, but to stupid to google it.
Who is a household name leveraging AI to upend an established market Every company. Are you stupid or what? They are automating their processes with AI.
No idea what you mean with "household", kitchenware? How the funk should I know anything about kitchenware/households?
Do you know Zeiss? The secret company behind ASLM? They subscribed 35k Gemini accounts.
And: that company is basically the sole single one company that produces the products they sell to ASLM.
In other words: they do not even have a competitor.
Nice insults, bud. So your point is people are using AI, but you can't name a famous example of a household name company that is upending established non-tech markets through AI? "Household name" means it's known by ordinary people. Tinder is a household name. Cloudflare isn't, despite having a 10x valuation. However, it's safe to say few outside of IT or investors who target technology could explain what CloudFlare does. And you name Zeiss...an optics manufacturer who is arugably a household name f
Re: (Score:1)
a household name company that is upending established non-tech markets through AI
No I can't.
I am not interested in stuff like this. If you are interested, why do you not google it? I am computer scientist in the move to become an AI engineer, I did not have the pleasure to work for a "household tools company" yet ...
And I guess you do not really know how big companies use AI: hint, they do not use it as search engine Ersatz.
Re: (Score:2)
The other issue is what is end game here with control and costs? For my curren
The reaaon this doesnt surprise me (Score:3)
If you que up a couple of bots? (Score:2)
You talk to coworkers or answer email. Or browse the web. But you can't increment the work until you get the results. The productivity gains come from anything you save time on. I've had spreadsheet operations run 20x what I could do by hand, but if you are just watching the gears spin while a couple of bots are running then there aren't as much productivity gains
Yea but tech bro millionaires! (Score:2)
Yea but have you considered how many filthy-greedy tech bro millionaires the LLM-bubble has produced??? Could the same bot-sitters have produced as many new gazillionaires, and an equal circular-investment bubble bordering on fraud??
I think not, so ha-HA!! Checkmate!
Botsitting _is_ the new work. (Score:5, Informative)
If the bot is 30x better than me on a bad day, botsitting is my new fucking main task. Obviously. In the last 6 months me and my AI metasubscription have grown to become a 10 head pro devteam with me at the helm. I've basically mutated into a chief senior lead and a full crew at zero extra cost and _ less_ effort for me. It would be irresponsible for me not to botsit and hold up everything by hand-coding myself. My current productivity would drop 10x instantly.
Bottom line: The bots are here and they've taken over. Get out of the way you slow-ass bipedal meatbag.
Re: (Score:2)
Bottom line: The bots are here and they've taken over. Get out of the way you slow-ass bipedal meatbag.
Yea, but they are not here yet, and won't be until they grow up and get their shit together. Until then you can keep them.
Headline is nonsense (Score:3)
Whether or not doing that "Botsitting" is better than doing the work myself does not change the above, and is a different discussion.
Re: (Score:2)
Stupid executives make stupid decisions (Score:3)
All of these executives who have been pushing for layoffs because "AI will replace the need for employees" are clearly incompetent at what they do. You do NOT replace good employees with an untested and unproven technology like AI. That's just common sense, and any employees who would trust the work of AI without verification of the results are also at fault.
AI at the moment, is good for certain tasks that will aid workers, but other than a BASIC use like front line for a phone queue to help route calls(and even this has been shown not to work acceptably in many cases) isn't a terrible use, but anything more complicated just requires too much monitoring and correction by humans to be worth replacing PEOPLE with AI at this point.
These executives have been scammed into believing that AI is ready to replace humans, but they show that they are incompetent at the basic things needed in management, the ability to decide what to do.
Re: (Score:3)
You do exactly this. The c suite (the ones making these decisions) are incentivized by triggering the stocks to move. They get that motion by any action that can move the stock price during their tenure. Will it leave a husk that is ready to implode in 4-5 years time? It doesn't matter... they did their job and it triggered their stock incentive payout and golden parachute. They are already gone and moved on to the next company where they can claim that under their reign the stock jumped 20% at company X.
Dev Ops example (Score:2)
Consistent with my observations.. (Score:4, Interesting)
Management has been pretty adamant that we are at the point where a developer should *never* look at source code, and work purely on prose. So some dutiful people have taken it to that extreme.
So they tell me they spend hours in planning, reviewing the resultant generated plan to give themselves confidence that the resultant code should make sense, then let it chew on making test cases, making code, running the tests, deciding whether the test cases were bad or the code was bad when failures happen, repeating until it has satisfied itself and then it lands in the builds that the still-human QA get a pass at. Then the human QA person often basically says "WTF, this is broken as hell" and the developer starts over incorporating the feedback from the tester into the flow.
Now the folks that discard that mandate and use CodeGen for code completion, more curated prompting, they get respectable speedups, but no hope for management to just toss those long term if they still need to curate the CodeGen. So management *really* wants the narative to work for the case above. They have raised the question if the real problem is the human QA, it passes all the CodeGen test cases and it's own code review, so maybe the human QA is just raising a stink to look relevant... Doesn't help that management understands neither the customer, software, or how to develop code, so they base things largely on wishful thinking on what the cheapest and most convenient answer would be that supports them getting big bonuses.
"work they couldn't explain if asked" (Score:3)
That will go well. They are delivering essentially unreviewed LLM output in what is now called "cognitive surrender".
Seriously, the amount of code that will have to be ripped out and thrown away because an LLM "wrote" it is going to be epic. It is pretty clear now that nobody that did this will come out ahead.
That does not mean LLMs are useless. They still work pretty well for "better search", but only with competent and mentally engaged human oversight.
my experience (Score:2)
I'm using AI for hobby coding projects using APIs and languages that I was never fully proficient in if at all. That works out great, even though I have to go through many iterations with the AI to get it right. If I was still programming like I was decades ago, for languages and APIs that I was expert in, it would be far faster to just code it myself.
So, that is what these are good for: non-experts can produce usable code that would require an expert without AI, but it's not production quality code. The
It's a promotion (Score:3)
Everybody is now a 'manager' of half a dozen AI agents.
Missing important info (Score:3)
Nothing new (Score:2)
I spend a lot of time babysitting my microwave too when I am heating something up... It still saves me time in the long run.