# Word of the Week: Botsitting and Botshitting

- **Episode:** [316](https://agents.truthliesandwork.com/episodes/316.md) (2026-07-07)
- **Type:** news-roundup
- **Tags:** ai, workplace-jargon, productivity
- **Time:** 7:21–15:14 (start 441s)
- **Listen:** [https://truthliesandwork.com/316?t=441](https://truthliesandwork.com/316?t=441)

---

[7:21] Al Elliott: Yeah, yeah, yeah. I feel like it's been at least a couple of weeks
[7:23] Leanne Elliott: Yeah, I have to, I have to thank Nicole Eisendorfer for sending this to me
[7:28] Al Elliott: Oh, she's a friend. We love her. Never actually met her. Never spoken to her
[7:31] Leanne Elliott: I've met her. I had coffee chat with her. Yeah, yeah, yeah. But we're, we're big pals on LinkedIn, aren't we?
[7:36] Al Elliott: Yeah, yeah. She was big friends with Bill Yost
[7:38] Leanne Elliott: Yes
[7:39] Al Elliott: Yeah
[7:39] Leanne Elliott: Yes, absolutely. So Nicole sent this to me. Thank you very much, Nicole. My new word is botsitting
[7:44] Al Elliott: I'm gonna have a go at this and I'm gonna describe it as sitting elbows on your desk, resting your chin on your hand, pencil or pen in your other hand, and just staring at your screen. And I think botsitting is when you go on ChatGPT, go on Claude, Other LLMs are available, and you type in something and you wait for it to come back and tell you what you want it to hear. So you're sitting there, just sitting there looking at a bot
[8:11] Leanne Elliott: Exactly
[8:12] Al Elliott: Is it?
[8:12] Leanne Elliott: 10 points to Gryffindor. Yeah, no, exactly. Botsitting is the labour of making AI usable. So feeding it content, supervising its output, and debugging its mistakes and cleaning up after it. So yeah, exactly. It's babysitting ChatGPT, other LLMs are available. So over the last few weeks, this has appeared in Forbes Australia, Business Insider, the Los Angeles Times, MSN, The Register, CIO, ITPro, the AI Daily Brief, and even discussions on Reddit and Hacker News
[8:44] Al Elliott: Wow
[8:45] Leanne Elliott: So the headlines are basically saying things like, workers are spending hours every week botsitting AI, the hidden cost of enterprise AI, et cetera, et cetera. So as I said, Nicole sent this to me, and the original post was from Rebecca Hines, PhD. She is a bestselling author and head of the Work AI Institute at Glean
[9:03] Al Elliott: What's Glean? While you're talking, I will, we will Google what Glean is. Sorry about that. What was her name?
[9:10] Leanne Elliott: Rebecca Hines
[9:11] Al Elliott: Rebecca Hines. Sorry about that, Rebecca. Oh, Glean is a coaching app that helps you lose weight and build healthier habits with a, with a plan that fits your lifestyle, apparently
[9:20] Leanne Elliott: Are you sure that's the correct glean?
[9:21] Al Elliott: It also, according to Merriam-Webster Dictionary, is the, it's to gather grain or other produce left by reapers. Um, yeah, or it could be the work AI platform connected to all your company data. That sounds more like it
[9:35] Leanne Elliott: Yeah, I think that sounds a bit more, a bit more on brand for a very good—
[9:40] Al Elliott: A bit more on brand. So yes, there we go. Glean
[9:42] Leanne Elliott: It's a cutting-edge research, research centre that partners with leading academic and industry experts to understand how AI is transforming organisations
[9:52] Al Elliott: Anyway, we're not sponsored by Glean. Why are we still talking about Glean?
[9:55] Leanne Elliott: We're not, but if you are interested, Glean
[9:58] Al Elliott: Yeah, so yeah, you can send us a cheque
[9:59] Leanne Elliott: HubSpot, if you're listening, not really
[10:01] Al Elliott: No, no, no, no
[10:02] Leanne Elliott: Joking, joking
[10:03] Al Elliott: Oh, it's just a bit of fun
[10:04] Leanne Elliott: It's just a bit of— just a little. Anyway, so this research comes from, uh, the Work AI Index, which surveyed 6,000 full-time knowledge workers across the US, UK, and Australia to understand what people were actually doing with AI at work. 6,000, quite a big sample. Quite a big sample
[10:21] Al Elliott: Also, I noticed you said surveyed, and I What's the difference between that and surveyed? Is it just intonation?
[10:26] Leanne Elliott: I always hear surveyed as like building
[10:30] Al Elliott: He surveyed his wondrous work as he sat back with a beer. That's God on the 6th day
[10:37] Leanne Elliott: But then I also say our instead of our. So I don't know
[10:40] Al Elliott: That's your northern roots
[10:42] Leanne Elliott: Exactly. Exactly. Surveyed or surveyed? You let us know. Let us know in the comments
[10:47] Al Elliott: If you're listening on Spotify, there are actually comments. We do listen to them and we do respond to them. We don't get many, but—
[10:54] Leanne Elliott: On Spotify
[10:54] Al Elliott: On Spotify, yeah
[10:55] Leanne Elliott: Yeah, yeah, of course. We have questions as well from time to time. Yeah, yeah. Anyway, the report found that 87% of people now use AI at work, and 75% say it makes them more productive, estimating it saves them around 11 hours every week. But then there's a catch. In the same survey, of course, only 19— sorry, only 13% say their organisation is actually performing better because of AI. So the researchers were kind of like, well, what's, what's going on? And basically their answer is botsitting. So when they mapped how people actually spend their AI time, they found 37% is spent botsitting
[11:32] Al Elliott: Wow
[11:33] Leanne Elliott: 36% is spending, uh, is spent generally using AI to get the work done. And the remaining 27% is spent learning how to use the AI tool in the first place. That works out to 6.4 hours every week spent supervising AI. It's basically a day
[11:49] Al Elliott: Yeah
[11:49] Leanne Elliott: Basically a day. The report went on to break it down even further. So on average, workers spend 2.3 hours every week feeding AI the right context, 2.2 hours checking its answers, and 1.7 hours debugging mistakes. And the researchers have even created what they call an exhaustion multiplier. So they found that for every 10% increase in the time people spent feeding AI context, They were 25% more likely to say AI was wearing them out. And that's when another new word comes in. I've got 2 words of the week. Are you ready? This is a swear. Anyone sensitive?
[12:28] Al Elliott: Okay
[12:28] Leanne Elliott: Ready?
[12:29] Al Elliott: Yes
[12:29] Leanne Elliott: Botshitting
[12:30] Al Elliott: Botshitting and botshitting
[12:36] Leanne Elliott: Yes
[12:36] Al Elliott: I love it
[12:37] Leanne Elliott: Guesses?
[12:38] Al Elliott: Botshitting. Oh, I don't know. Is it, is it just where it comes, it comes out with a load of shit?
[12:43] Leanne Elliott: Kind of, yeah. So if botsitting is kind of checking the AI, then botshitting is basically pushing out AI-generated work without checking it
[12:51] Al Elliott: Right, like slop
[12:52] Leanne Elliott: Yeah. So according to reports, 69% of workers have admitted they are at times a botshitter
[12:59] Al Elliott: Really?
[13:00] Leanne Elliott: Yeah
[13:01] Al Elliott: So just so I'm clear, so botsitting will be sitting there looking at the, looking at the, what's being generated on the screen, then most people We'll go and check that and then send it. But actually, 7 out of 10 people have just gone Ctrl+C, Ctrl+V, email, pub
[13:18] Leanne Elliott: Yes, exactly. Exactly. And perhaps the worrying statistic, obviously this from a productivity perspective, not great anyway, but from a people perspective, what I found particularly concerning is that the researchers found that frequent botsitters So people who spent 40% or more of their AI time supervising the systems, those employees were 73% more likely to be actively looking for another job
[13:47] Al Elliott: Oh
[13:47] Leanne Elliott: I know
[13:48] Al Elliott: That's so— why are they looking for other jobs? Is it just because it's boring to sit there and botsit?
[13:54] Leanne Elliott: Yeah. Yeah, exactly. Exactly. It's all just a bit dumb
[14:00] Al Elliott: At the moment, I don't see any other way around it. Because we— it doesn't know our work well enough. I think it'll be in a year's time. It will. It'll, you'll, you'll, it'll know you well enough so you can just set it off and go to do its own thing. In another thing, right? So I've been using ChatGPT to create 3D renderings of our house. And it started putting an extra wing on our house in the 3D render. And I must have spent 25 minutes going back and forth trying to get rid of this, this render, this extra, extra wing on my render. And then I got really annoyed with it. And I was like, well, screw it. Screw you then
[14:30] Leanne Elliott: It's funny how quickly that happens, isn't it? How like one week you're blown away and the next you're losing your rag because it's like, just be better. Just be better, chat
[14:39] Al Elliott: Do magic better. Anyway, so, uh, so any, any solutions to this? Any conclusions?
[14:45] Leanne Elliott: Well, the report did say— it is a really good report. I will leave a link to it in, in the show notes. It did, um, it finishes executive summary by saying there's plenty we still don't know about how AI will transform work. But this much is clear: organisations must build the human infrastructure, not just the technology infrastructure, that makes AI worth using, or they'll keep paying the bill in botsitting, in botshitting, and in the excess of people who get fed up cleaning up after the bots. I've seen this week
