The Truth About Corporate AI Spending (It's Not What You Think)

So Accenture, they found out that most of their— I think they're averaging like $1,000 per head in terms of AI spend. So they're limiting employee AI spend to $29 per month per employee. What do you think about that?

I think it's really smart. And there's 2 reasons. One, $1,000 really adds up. I'm assuming you're talking about $1,000 a year, not a month, right?

I think it's a month.

You sure? $1,000 a month is a lot.

Some of it was going up to a month. So I don't think they were saying everyone was, but because they saw some employees going up to like $1,000+ a month, they're setting the limits on it. And let me pull up this post for you so you can actually see it. So let me share my screen. All right, Neil, do you see this?

You're right. $1,000 monthly. Tuck me in.

Yep.

And now $29 a month per person.

Yeah, so you've been promoted to consumer. Okay, go ahead, continue.

I think it's smart. Even in my organization, I've heard direct hand from engineers that I trust and that get poached from other bigger corporations that they're just like, a lot of these models are just expensive and some of the older ones do just as well. And I don't know why people need all these advanced models to do basic stuff, 'cause that's the reality. Most people are paying for the most advanced models to do the most basic stuff, and it doesn't make sense. And then the next thing is, is there's so many open-source solutions now that can get a lot done for free. You don't need to end up spending an arm and a leg. Yeah, you may need a better PC or computer or setup, but that's much more palatable for corporations than it is to spend $1,000 a month.

Dude, you know what I'm just thinking about right now? Like, because Anthropic and OpenAI haven't IPO'd yet, right? So when companies really figure out how to optimize this, I wonder how it affects them down the road. But that's a completely different conversation. So,

uh, they released a revenue recently, right? You saw the OpenAI revenue numbers?

$16 billion on Anthropic Q3, I think. Yeah.

Correct.

Which is—

and then they're gonna get to $100 run rate.

Yeah.

But the, the interesting thing when they call revenue, revenue is not always all revenue. Some of these players have deals with other providers who provide them a lot of the infrastructure or money or whatever it may be, where they're getting a percentage of the revenue. So it's like eBay. eBay, uh, Eric may sell some Pokémon cards on eBay. Okay. And these Pokémon cards that he had since he was a kid may go for $1,000. eBay's revenue from that is not $1,000, even though that product sold on eBay for $1,000.

It's GMV.

Yes. eBay's stake or percentage that they may keep may be 5%, I'm making up. I don't know what eBay actually takes. So let's call it $5. eBay's real revenue is $5 per transaction or for that transaction. And it's the way these guys are reporting numbers and tracking them and putting it out to market varies a lot because they're not public. They can do whatever they want, but the moment they're public, people will look at how much is your real revenue and what are you collecting?

Well, actually, I mean, again, we don't have to go deeper into it, but I think I was reading something in passing where Anthropic's gross margins are like 80, 85%, but then, you know, operating revenue is like 11% or something. But let's go back to the $1,000 real quick. So I think there's caveats for this. Okay. So again, Accenture is saying, okay, they're seeing some employees spending $1,000 a month, which is a lot of money, and they're converting everyone to $29 a month. Now, Ramp, we're talking about Ramp earlier. So Ramp actually has an AI spend index. Okay. So the median spend is $11. Okay. So they're saying, okay, we're going to give you like 3x the median spend, which is like if you do the math on it, it's like, okay, that's pretty good, right? And so I think as long as you're slightly above the median spend and then, but you're not limiting your best people, maybe your best, as long as your best people can justify and they can prove that they're driving revenue or they're saving money on it, then great. Right. But there has to be a process for it because you can't just let everyone YOLO and do whatever they want. Now, the cool thing about Ramp too, Neil, this is not a Ramp sponsorship, but the cool thing is you can connect your accounts and then you can see how much people are spending on average in your company, who's spending the most. And then I think you just need to have conversations with those people. So I don't think it's like a, I think it's cool to blanket, like, set a limit for everybody, but then the people who do want to take it to the next level and are proving value, you should let them do it. So I think it's a yes-and situation.

Yeah, I agree with you. And even though Accenture is saying it's limited to $29, if someone would bet me $1 million, I would bet that Accenture is not limiting every single employee to $29. And there's some people who can get exceptions if you wanted to go over, Like engineers or people like that, because they must have some engineers doing some projects and they're okay with them spending more. But just as a general corporate policy, they're just saying, hey, $29 is your limit. And then for certain departments, certain people and roles, they have no choice but to spend more. And that's

okay. I agree with that. There needs to be a corporate policy for these things because remember what I just said earlier, if you have 6 billion people in the world that are just barely using this stuff and you have 1 to 2 billion that are using it as a glorified chatbot, What do you think is going to happen in a company like, like Accenture where they have, you know, tens of thousands of employees? The vast majority of people are not going to be using it in a smart way, and that's going to be costing the company too much money. But there's going to be 0.2% of people that are using it in a smart way. You let those people spend more. All right. So I wanted to take a moment to tell you about my podcast co-host Neil's agency called NP Digital, and they work with a whole host of global companies or global organizations. Also, Neil has SEO tools such as Ubersuggest and AnswerThePublic. All you have to do is go to npdigital.com to learn more, and we'll see you on the other side.

Dude, did I tell you I downgraded my ChatGPT subscription?

Your own?

Yeah, my own. I downgraded it.

I love that you spent the time to actually do that. Oh, how much are you paying a month now?

Uh, I don't know how much I'm paying. I believe I was paying around $200 a month before. I could be off on the number, but I'm pretty sure it was somewhere around $200 because I would always get an email from my assistant being like, Neil, can you go get me the receipt? They didn't Email it to you. We need it for accounting purposes. I was like, all right. And then I downgraded. I'm like, man, I'm like, I now really only use ChatGPT as a glorified search to do like research and stuff. And I'm like, why the heck am I on the highest plan for this? So I downgraded it. I think it's like $20 a month or $29 or something like that. I could be off on the pricing, but I know it's not $50 a month and I know it's not free. It's that pricing that's in between there, but it's not the highest plan.

Dude, I love my ChatGPT now. I was driving to see my dad last week, right? So when you drive to Arcadia, it's good, it's like 45-minute drive, 40 to 45 minutes. I just called Doodoo and I was just talking for 45 minutes and it's like, okay, what about this? What about this? Okay, can you move this forward over here? Can you advance this over here? This is like, we were just going back and forth.

Isn't Doodoo your muse?

They're both called Doodoo. So Doodoo is muse and Doodoo is also my dot too. They're both called Doodoo. So Uh, anyway, so that's my OpenAI bot. And like, it, it's really good because I'm like, oh, we're trying to make these ChatGPT plugins right now. Okay, we're trying to do this redesign over here. We're trying to do this over here. Oh, can you hand this over to Sean over here? Can you do this? Right? So I'm like, I'm just going back and forth with it. It's like you just talking to like an employee on your team for 45 minutes and it's just getting stuff done. So, um, that to me is worth it. But I'm not on the $500 plan. I'm only on the $200 plan.

There's a $500 plan?

They just added a $500 plan. Yeah.

Wow. Yeah, I should actually check to see what plan I'm on. I know it's the cheaper plan.

I'm sure yours is $29.

I'm on Plus. What's Plus?

$100, I think?

Plus, $20 a month.

Uh-oh. Well, yeah, they brought it down a little bit.

Yeah, this is great. Let's see all plans. Oh yeah, you're right. So, oh, they have an $8 a month plan? Expand access?

No, at a certain point, saving money is not worth it. That's diminishing returns.

Eric knows me really well. I'm like, ooh, I can save even more.

Okay, so this guy, how do you pronounce his name, Neil?

Sujayth? Sujayth? Yeah, I have no idea, but it sounds right.

Okay. All right.

I should know how to say his name, but I don't.

Yeah, you should know how to say it. So this guy says, I got a lot of pushback for saying that one product manager and one engineer can now beat an entire Now, Dan Schipper has a name for it, the pirate and the architect, right? So it used to be— remember Amazon's Jeff Bezos? He used to say, uh, 2-pizza teams, right? We want to have 2-pizza teams. Now today it's 2-slice teams, okay? So you can only get 2 slices for the team. So let me play this video real quick.

I've started to come to is you really want to use very, very, very small teams. For a long time, the standard for team size was the 2-pizza team. So that's like 8 to 10 people. That's what Jeff Bezos came to. I think in the AI age, I call it a 2-slice team. Like, you want 1 or 2 people max. You can get so far so fast with 1 or 2 people that anything more, there's a lot of coordination overhead and there's differing visions and it just doesn't work. 1 or 2 people is really, really great. And I think the composition that works for us, and I think it's starting to work for other people, is I call them pirates and architects. So one person on the team is a pirate. They are slop cannons who are absolutely obsessed with finding value. That's me. Like, I'm just out there just doing lots and lots of stuff and I'm throwing it away and it's all messy, but I'm going to find something really interesting. And then architects are people who are like looking at a messy system, like maybe something that I built that's completely hardcoded, and trying to help shape it into something that's valuable and beautiful and extensible. Pairing those types of people together, I think, is very powerful in AI.

I think this is interesting, Neil, because we have our agent builders team right now, and then everyone has their own project. Most people are just working on their own on a project. Maybe there's someone else paired with them, but we have one team where there's 4 or 5 people on it. That team is the slowest one because there's so much drag that comes from coordination. You know, if you have 5 ideas, I have 5 ideas, someone else has 5 ideas, and then Noah has 5 ideas too, that's 20 ideas. And everyone wants to get their ideas forward, right? So I like this whole idea of a pirate and an architect because I'm probably more of a pirate where I just like throwing spaghetti against the wall. And then you do need an architect that can help you shape this stuff into like clay into like a little castle or something like that. But I do, I have found firsthand experience that when you have You start to have more than 2 people building these things, it starts to get really complex, and there's politics start to build even at such a small size. So,

dude, I'm, I'm with you. Uh, I do like the smaller teams. I was talking to one of my team members the other day, they're like, you used to run Ubersuggest back in the day, how'd you do it? I was like, with very few people. Fewer people means faster deployment. I do think in Product and engineering and design, a lot of that stuff is definitely true. I do believe when you start getting into complex situations like finance and banking, small teams don't work out. Like Bank of America found out that through Claude that there's tons of bugs, so they hired 1,000 contractors to fix all the security bugs. You need a big team and you better crank that out really fast and fix all those bugs, especially because it's related to finance. But on the flip side, for some business cases, when you're doing, let's say, a merger and acquisition and you're trying to buy a company, you can't buy a company with a 2-person team. And I know the example there, they were not saying for M&A, but you'll need a finance person, you'll need a legal person, you need a person who actually knows the business really well, someone who understands maybe account management if it's in marketing, someone who understands You know, just looking at their delivery and seeing if they actually have a really good product or service, the churn rate, right? That gives them some data on that as well. Someone in sales to analyze their sales process and if they do a good job, because then you can figure out where to fix a lot of this stuff. But that usually requires more than a 2-person team. But I know in that video they're mainly talking about, you know, building products.

Yeah, I think it depends, right? Like earlier we're just talking about Meta having 200 people on that super AI team, by the way. They have 70,000 people. 200 divided by 70,000 is 0.002857. I have my calculator out.

They have how many people on the team?

200 people on the AI team. 200 people out of 70,000 employees total at Meta, right? And then even the way Mark Zuckerberg was talking about it, it's like, yeah, it's like this elite team and they just want to move quickly. And to your point, like 200 of 70,000 is really not that much, but that's what it takes to move lightning fast, right? And he wants that team to move lightning fast. So there's that. But then I wanted to bring you up another thing over here because we're talking about private equity. Okay, so Bending Spoons, would you say Bending Spoons is kind of more like a private equity type model?

It is purely a private equity type model. Hence Berkshire Hathaway popped in money and then the stock popped like 40, 50% in one day.

So check this out. So Bending Spoons, okay, I don't know if you saw this, but this guy Federico who works at Bending Spoons, they're based in Milan. So, you know, Italian guy, right? So today we're laying off a large part of the Airtable team. So remember, they bought Airtable, and I think they got a good, like a decent deal on it. So what Bending Spoon does is they bought Evernote, they bought Airtable. What else did they buy, Neil? Do you remember? Oh, Vimeo,

Evernote, Eventbrite, Evernote. They're on the right side.

WeTransfer. Yeah. Meetup. Yeah. So they buy these struggling tech companies from the past. They used to be iconic tech companies, right? Like Eventbrite, Meetup, AOL, Evernote. Right. So he's like, look, today we're laying off a large part of the Airtable team. This is not a reflection of their work, but a reflection of our deep belief that small teams ship more and better. For Airtable, better is about to look like a new core team shipping your top requests monthly, upgraded, you know, this and that, new performance and infra teams. Okay, to everyone leaving, thank you. If you're hiring, these are excellent people and worth the call. So before you react, Neil, let me tell you why this is such a tone-deaf, uh, You know, posts. Because instead of like— I even posted here, I was like, I was like, dude, where are all these people that you're laying off? Like, give me a list of those people, right? Um, and but it's like, it's just like, yeah, you know, these are excellent people and worth a call. It's like, how do I know who to call? Um, and this is like, why are you even announcing your new product features when you're talking about a layoff? Like, those should be separate things.

Totally agree. I'm like, this is a terrible message. I don't know who wrote this. I'm assuming they're at Bending Spoons.

She leads product at Bending Spoons.

I just looked up how many employees Airtable has. This is from a Google overview. It says 800 or so. I'm guessing they're laying off more than 50% is my guess. I don't have any stats or data, but when they say a large part, Bending Spoons model is you gut, you increase profit. This is what private equity does. And this is why people say private equity is not the place where this is where companies go and their innovation dies. And I really do agree with that, you know, and people always talk to me about innovating and we have a lot of people on our team who try to innovate. And my co-founder and I are like, so if you're telling me this is going to be a new thing, is anyone doing that? And usually because of AI, people move so fast these days. They're like, yeah, there's these other companies that are doing it. They're really tiny. And my co-founder and I usually just shoot an email and it's like, let's just see if one of these companies will sell to us. It'll be easier.

Yep.

By the way, that is not where you go to innovate.

So let's call something out real quick. Okay. Evernote back in the day, right? So Evernote used to be like the number one note-taking company. Okay. So Evernote's peak headcount was roughly 433 employees. Guess how many people are running Evernote today?

30.

Exactly 30. Okay.

So that's 70%.

They cut, they cut 93% of staff. Okay. And they're doing— that is their playbook, right? Airtable. Like, by the way, the core Airtable team, they're all on HyperAgent now. So they spun that off from Airtable and they sold Airtable, right? So the core Airtable team, they're already in HyperAgent. They've already sailed off to somewhere else. So they're okay with, with, you know, cutting the rest. Um, so anyway, that's the model. Cut 90, 95% or so, just have 20 to 30 people running it. And then the thing, the way they run Bending Spoons is they have, um, different product managers, different engineers. They're kind of working on different projects and they're just kind of moving around. Um, so it's, it's interesting for them, you know. Yeah.

Yeah. What's interesting as well is when you're publicly traded, most people can't cut too deep. Okay. But their model in the public markets, know this, is buy companies that aren't high-flyer growth companies anymore and just gut them. And then you just take a cash flow pay model, right? Money is money. People are happy with just cash flow. If you're publicly traded and let's say you're focusing on something that's core, like a Microsoft, a Google, Atlassian, whatever it may be, If you announce that you're cutting 95% or 90% or 80%, the market would be like, something's wrong, and it can actually hurt your stock price more. Bending spoons is in a different model. The market accepts it and they're like, oh yeah, this is great. Cut, cut the fat. Dude, this is what we—

this is interesting. Neil's right. So Berkshire put $330 million into it yesterday, so sparked a share rally.

So

by the way, year

to date, put it a while ago.

Reuters said it was one day ago. So this is probably some wrong Google overview or whatever, but market cap $32 billion. They're up 20%. 25% in the last 6 months or so. So not bad for what they are. Yeah. Are you sure, Neil? Here, let me show this to you. Like, either I'm wrong or—

October 13th. It's October 13th.

Oh, 3 days ago. 3 days ago. Yeah. So this is just like later news. Bending Spoons skyrockets 41% as Berkshire Hathaway makes $330 million. Okay, I see. Oh, and this was from their CEO, Abel, Greg Abel, right? So

yeah. Yeah. The reason I know the dates is a lot of times when I see this stuff, I'm like, man, You know, I, I, no one knew that they were going to do it or else it would be insider trading. I then go and look at the options chain and I'm like, well, if I put in money at this price, how much would I have made? I was just curious, right? So it was pretty good.

Anyway, all right guys, so that is it for today. Um, we hope you have a great day and we will see you tomorrow.