AI usage costs are destroying SaaS business models

0:12lady. Okay, [laughter] don't do that. [clears throat] Ladies and gentlemen, welcome to the factory floor. Today I am joined by guess who? The same two nerds I'm always joined by, uh, Zach and Corey. Uh today we're going to be talking about um once again AI and how it's infiltrated all of our minds and brains lives of course uh but specifically this time in the context of SAS pricing. So obviously we work with a lot of SAS companies. We talk to a lot of people in SAS and so how do you price your product? How should pricing exist in the world of AI? So that is what we are going to be talking about today. Does that sound good to you guys? >> No, I think we should change the topic.

1:00>> Let's change it. Let's do something else. [laughter] [gasps] >> Okay, let's do it. So, uh Corey, why don't you like just get us going? Kind of give us like a broad overview maybe of, you know, you're the you're the marketing guru at the very uh top. So, what what do you see is the biggest place where AI is actually changing the way SAS pricing works? Mhm. Okay. Here's how I think about it is that first of all, I think a lot of people have been like, you know, it's like we're all like not interested and we're kind of like we're all a little bit burnt out about how AI is changing everything, but like we have to know because otherwise you're going to get majorly left behind or screwed over. We're going to end up

1:45shooting yourself in the foot. So, like this is one of the things I've been thinking about where it's like, hm, I don't know what the strategy is here, per se, but this is something that everyone has to be hyper aware of and that everyone's for sure thinking about right now because we're basically forced to to make some of these changes that I'm going to talk about. And it's not going to be the whole like, oh, SAS pricing now is um all based on like results and how it's all just about like uh like success-based pricing. I don't actually think that that's necessarily the point. I think that's one of the points. We'll talk about kind of the nuances of it, but here's the deal. We had software 1.0, you know, which was

2:29based off of on premise installations on your computer. And how they tracked that was per license, right? So you built software, you licensed a version, people pay for that, they install on their computer, they can use it for as long as they want on their computer, right? And then we have this thing called >> old Adobe old Adobe users, right? >> Yeah. Yeah. [laughter] Exactly. Adobe, Word, Excel, um even like Salesforce, Netswuite, um was it Netswuite or I was >> Yeah, I think Quickbooks, right, was our original onrem kind of software. So, uh all of those were the pricing was built around fundamentally how people used programs on their computer, which was that everything was installed locally.

3:20So, when you want a new version of software, you pay for the next version. you know, there's like word one, word two, word three, word four, so on and so forth. And that was actually a pretty good pricing model because they basically built software around having like these annual upgrades that they want you to do essentially and pack it full of new features and upgrades, things like that. Then we had this thing come around called the cloud, right? And this is basically web- based apps, software as a service where now there's no longer versions because people are shipping updates daily, sometimes even hourly. So, how do you charge for a license? And how do you charge for something that people are constantly

4:03using on their browser, possibly even on different devices, different computers? Well, now I have the idea of subscriptions, right? So, you essentially pay for access. It's like leasing software. Now, so that's like software 2.0. Now, I think we're entering into kind of already entered into an era, but we're all like figuring it out. We real time where it's really the software 3.0 where we've even gone beyond like the cloud and access based software. It's a lot more outcomebased, resultsbased, and it's based off of like computing power because now we're all renting software from Amazon, from uh Nvidia, because everything is powered by AI, which is like really, really computationally heavy. Um, and so again,

4:54the pricing has to shift again because we're not just paying for access of software. Now we're paying for how much that software is doing for us because AI is expensive. So does that kind of like paint the picture a little bit about like >> oh yeah >> 30,000 foot view last 30 years basically. >> It's and it's less like we're not necessarily saying like okay this is how it all works now. It's more like, hey, this is the the change that's taking place and we're still getting our footing or the world of SAS is still getting its footing on how pricing should work in this new world. Like is it credits? Is it you know what? There's multiple different kind of avenues and we're still kind of getting our footing

5:36with this new uh SAS pricing 3 or software pricing three. >> Yes. And and the main difference is that in the 2.0 know world the software that you paid for was running all of its own compute right so whatever you do on there is powered by the servers that they pay for in software 3.0 you know world now you're not just paying for access to that software and that compute but also the compute that it's paying for and using from other platforms all the AI providers right openai anthropic um >> Gemini well Google right uh so on and so forth right there's like five you know Grock is one of them too um so they have costs that they're paying for they're passing on to you and So now we have this kind of mixture of two different

6:28layers of computing that you're paying for, not just one. >> Okay. Okay. I think that's making sense. Okay. So then with the new the new development, the new version three, we'll call it version three. >> What is like the stuff from version two that doesn't fit at all to version three? Like there's this doesn't work at all in in V3. >> Yeah. Well, one of the big things I've been noticing is Fremium actually, which is crazy because Fremium was such a huge innovation for software 2.0 where people figured out that the computing cost for a free user was so low that they were like, let's just give it for free and then we'll have some kind of cap, some kind of payw wall at some point that'll get people to upgrade. And that'll be a

7:16huge advantage for us both from a product and a marketing perspective because it really takes the try before you buy mantra to like the ultimate extreme. >> The problem now with fremium is that the computing cost is no longer low potentially, right? The computing cost is high because the computing cost is low for your own computing power but not for the AI computing power that whatever software you're paying for is buying from the AI providers. So it could potentially be extremely expensive actually right imagine you have Canva where it's like hey free access to our little you know graphic design software but then anytime you use an AI generation for graphics it could cost them dollars per generation

8:08>> right now this is can you draw a distinction here between premium and free trials because I think that'd be a like I don't see free trials dying but at premium is going away. >> No, actually they're not so dissimilar because if you think about it, someone could come into a free trial and also really abuse uh AI features or AI usage that could rack up a bill of, you know, tens or even hundreds of dollars um if it's uncapped. And that's kind of the whole point of why fremium needs to be adjusted and is no longer the same as it will ever be is because anything that is uncapped when you're using AI is going to be expensive until potentially the AI becomes a lot lower cost. The computing power is super low now. And

9:02we're kind of seeing that where a lot of the AI providers will make their older um their older models and their like mini models quote unquote a lot cheaper. >> Yeah. >> But it's also not really they're not really incentivized to like keep a lot of those around because they're a lot cheaper. They want you to use the newer, bigger, beefier models that cost more. Right. Well, and I'm sure it just doesn't make sense for them to like to tie in your software product with a an antiquated model either, like an antiquated AI model, you know, like why why would you do that? Have and have them bifurcated that way. It would just be far more complicated. It's more to explain in your >> comparisons of you like what does a free

9:48trial constitute or premium constitute? It's like well you get the old models like but it's and people have an even harder time delineating between that more confusion, >> right? >> Yeah. and then they're not really getting the value possibly that's going to trigger them wanting to make an upgrade because now they're like, "Oh, did I really experience the full power of this platform? Was it really a free trial or was it a limited handicapped nerfed trial?" Right? Um now there's one other caveat to this that uh this could also no longer be true if and once software companies start to figure out how to build and run their own models because then again the computing power once they bring that inhouse they can

10:31offer an open-source model for much cheaper than the AI providers because they have all the control and all the compute um which is going to be you know basically the difference of like manufacture ure costs versus retail costs. Think about it that way, right? Like >> like the consumer is paying the real retail cost of AI, the SAS company's paying the wholesale costs, but if they can get down to the manufacturer costs, then it's going to be a lot cheaper. >> That's crazy. The whole like framework is is kind of like wild how you could think of it just like any other product is being developed. Do you think that that could is there an opportunity here for feature gating where like you know so say somebody has

11:15AI as part of a like maybe their their middle Goldilocks offer and then their premium offer or maybe like enterprise but then the starting one doesn't have any of it and you can start them on a free trial where they get access to all the AI tools and get to try them out for two weeks and then they are pushed back to that base model or base plan that could be free, could be a premium one, but doesn't have any of the stuff that they were just playing with and that actually showed them the value. It sounds like that might still be a thing, but >> yeah. Yeah. Yeah, I think so. There there's definitely we're seeing with this with uh with notion for example where there you basically every SAS company now has to adopt two

12:04simultaneous pricing models. One is the access for the software. Two is the usage of AI features >> whether they're providing those in-house more at the manufacturer level where they're running their own models and their own compute or if they're buying and uh outsourcing all that compute to an open AI or an anthropic. um there will always now be some kind of usage based pricing tacked on top especially for AI >> and again this goes back to well how do people use the software and where do the costs come from usage based pricing is essentially going to always be around now or at least be a factor of any pricing model because that's part of the cost because when they're buying it from OpenAI or anthropic, they're paying for

12:58a certain number of tokens, right? >> And so >> in the past, you had a provider like AWS or Heroku, for example, where there are like these managed compute solutions. You don't have to buy your own servers. You can just essentially lease from the cloud and they're going to give you some kind of computing power for really cheap that gets you the bandwidth you need at a very like stable kind of rate. So every month you pay $10, you get two servers and that can handle x amount of users. And so the the costs were fixed. Now the costs are always variable for AI where it's every single token you pay for every single usage, every single API ping essentially. >> So we kind of see like now there's going

13:45to be >> there's going to be SAS pricing and then there's going to be AI pricing >> and they're going to be bundled together. Do you think that that matters like having the So, for example, I just signed up for Weevi, which is one of the node-based um design editors that Figma just acquired. And their pricing, if I'm not mistaken, you do have a free trial where you get access to a certain number of credits and then you get access to I think that it's their I'm pretty sure it's their most popular version where it has just the right amount of features and, you know, models, access to all the things that you would want in using that tool. And it came and it came with a set number of credits. Now, if I wanted to

14:29pay for that software, I'm pretty sure they have a baseline tier, which is their lowest cost, plus, you know, x amount of AI credits inside of there. And then there's like a a credit cap on the lower tier. And then in the middle tier, you have another amount of credits, but you can go a little bit higher and pick how many you need and that adjusts your costs as well. >> Or, you know, you have the enterprise plan, which is, you know, gang busters. you know, you get get everything that you want because you're paying us so much money. >> So, do you think that it's are you seeing anything with like the the way that those >> it's almost like usage gating or usage tiering in addition to the >> the credits? Um, or is it do you think

15:14that there's some utility of just having your baseline software costs and then on top of all of them you have that other layer of Oh, by the way, if you want AI, you can just add as much as you want regardless of what tier you're on. You're not gated based on how much you're paying for your base software package, right? Yeah. I think they're probably going to be mostly unrelated where you can pay for any tier that gets you access to a number of users, features, capabilities um within your own platform, >> but then any of the AI enabled features or usage is going to be sort of a separate system. I think a lot of people right now are going with credits and some sort of credit package. And that's

15:59again also just how OpenAI and Enthropic have been charging is they're like, "Hey, you sort of like top off, you know, you can buy you can spend $100 per month and that gets you these credits that don't expire. Anything over that we charge you at some sort of usage. If you want to buy bigger packs in advance, you get some volume discounting, right? And then the idea of a credit is flexible enough to where you can make one credit equal five API calls or you can make 100 credits equal two API calls. It's sort of like up to you and how you're using the AI to figure out how to like pass that along to your customers um versus just copy and pasting what you're being charged. So like well because there were

16:44some things like so generating an image with GPT um image was way less credits than generating something with V3 which is the Google video creator right so that's how they're having you buy those base amounts of credits so that you can they're hedging you know your your costs like if you're going to be doing a lot of videos you're going to need a lot more credits but if you're just doing GPT image gen then >> you don't need as many probably. Yeah, >> it's like this very strange like it reminds me of um what is it in like you can buy credits in Fortnite. What's that called? V-Bucks. >> V-Bucks. >> You know what I'm talking about? That to me that's that's the itch that's being scratched is like you could buy what

17:24a,000 V-Bucks and it's $22 and you're like how many dollars is a V-Buck? And the same thing happens with these credit. Like how many dollars is a credit? And okay, so then they're API calls. How many API calls fit into a credit? So, how many dollars is an API call? And then like how many API calls do does each user use? It's like this whole like labyrinth and weird like um exchange rate to deal with and it's honestly kind of annoying. >> Yeah. Well, you want to go even deeper down that rabbit hole? >> Yeah, I do. It's because how LLMs work is they essentially tokenize a word and they have this whole crazy system that, you know, converts everything down into binary to where you can essentially turn a word into a

18:12series of numbers. And that's kind of what the tokenization process is. And so when you're buying tokens from OpenAI, you know, a thousand tokens could be 500 words or it could be 200 words or it could be 800 words just depending on the word and the word length and the >> the sentence, right? The combination of sentences. >> That's unbelievable. >> So the bigger your words and the more eloquent you are, the more you're going to pay Chad GPT. [laughter] >> Yeah, pretty much. So you got you got to ask dumber questions to the AI robots be like make rich please and it will it will sum [laughter] it down and cost less tokens. >> Well that kind of is part of the magic of prompt engineering and what a lot of

18:56people have realized with coding especially is you have a a context window and a context window is essentially the short-term memory of the AI in how much you can give it and then how much it can spit out. It's sort of like if you told me, "Hey, repeat after me." And then you told me like I can only say so many things before like I starts to get fuzzy and I start hallucinating maybe like what you what I'm supposed to repeat back. >> Yeah. And so people have realized, oh, instead of saying, make me an image that is photorealistic and X Y and Z, people just say image colon hyper photorealistic detail, like they just kind of mash everything into this like >> keyword blob >> because it uses less tokens.

19:45>> That's hilarious. So Kevin Malone fares better with chatbt than Robert California. [laughter] >> Why use slot word when fewer do trick? That's exactly the AI mantra. >> That's what they've been teaching us this whole time. Okay. So, I have like a larger question based on this token and and usage, pricing, and all that stuff. Um, is the AI feature revamp of software making SAS companies a lot less money or a lot more money? Like I is is profitability going down because they're like sure they had they added this cool AI feature and maybe they got some new customers and maybe they could charge a little bit more but the cost is just like the the scales are tipping in the wrong direction.

20:32>> Mhm. Yeah, that's a really good question because we look at companies like Cursor for example. Why is Cursor raised billions of dollars? It's because they subsidized the cost of the AI compute that they were paying for that they were passing along those savings to their customers. Because I remember even like a year ago in cursor, I was just vibe coding like crazy cuz I was like, "Okay, they're going to charge me $20 per month and based on how much I'm using all the AI, if I were to pay OpenAI directly for this, it would be like hundreds of dollars." OpenAI is still getting paid. Cursor was just subsidizing it for me. they were eating the cost. So, they can't do that forever, right? And it's

21:16sort of like back in the day when a lift was $5 for an Uber was, you know, $3 go on the block and now it's like $20 to just go down the street because VC money was subsidizing it. Well, like you can't change the the unit economics, right? At the end of the day, OpenAI is going to get paid. At the end of the day, actually Amazon and Nvidia and Oracle and whoever their like real computing uh partners are like they're going to get paid. Um and so right now AI is making startups a lot less profitable. um until they start adapting to this needed change in their pricing model and sort of having this you know manufacturer level, wholesale level, retail level pricing of AI and how they use that or pass that along to their customers. Um,

22:10in in the long run, it totally can make SAS companies a lot more money because instead of being maybe capped at $100 per month possibly, now it can be $100 per month plus $200 in AI usage, right? And then maybe of that $200 in AI usage, they have a 50% profit margin, for example. So they just doubled the their kind of like take-home, you know, profitability per customer by introducing this AI usage model on top of it. And it could be, you know, completely optional per customer, but it it definitely has an opportunity for every single startups to increase their average revenue per user and even their profit per user.

22:58>> Dang, that's sick, dude. It makes well it makes sense that it's um that it's costing them a lot of money now but you know with all the things it's going to eventually become more cost effective and it's just a matter of time like can you hold out and I think that by making those shifts and having the flexibility within your pricing that you'll be able to weather the storm a lot >> a lot better. I am curious. So, what do you think? You've you've mentioned outcomebased pricing before, but I am curious. I'm curious how that's going to go, especially when you have a lot of, you know, the marketing around software always almost always comes down to increasing ARR or increasing MR. [snorts] >> So, what happens when that train of

23:47thought goes all the way down? Shout out to our phones of friends at Fletch, Anthony Perry and Robert Kaminsky for highlighting that. >> Yeah. Yeah. Exactly. I mean, that's definitely a big risk. It's one of the other reasons why I don't think that SAS pricing is going to move completely to outcomebased pricing or resultsbased or success based, however you want to call it. that whole idea just being, oh, instead of charging $100 per month, I charge $1 per valuable thing that I deliver to my customers, right? And then maybe on average they do a hundred of those per month, but now it's a lot more flexible. I still don't think that's going to be the case. There have been companies like um Intercom's Finn agent,

24:32for example, that have like >> really moved to that model completely. It's like a really big bet and I think that that can work when it can be a cost savings to your customers. Intercom was super expensive. Like let's be honest, it was it was like dreadfully one of the most expensive uh customer support tools out there. Um and so people were like, "Oh, I only pay for, you know, completed successful uh support tickets. Cool. that can save me money, but when it starts costing people more money, they're going to be like, uh, maybe we should go back to SAS because it was kind of nice being able to predict all of our costs. I don't think that there's a world where anyone is ever completely comfortable with 100% based, usage based

25:21pricing. Um, and that's really what outcomebased, success-based, resultsbased pricing is. I think that the future and really what people are going to adapt the most is going to be this combination of SAS plus usage and maybe your base SAS kind of gets you a certain number of credits or a certain number of usage and then anything above that you pay >> either in credits or tiers or uh or just like a pure kind of flat rate for whatever usage you uh >> yeah usage metric you want to measure. But um yeah, I I don't think that it's going to go 100% to outcomebased pricing. So I think MR is still safe. It's just going to be like MR plus >> and [snorts] I don't know how we going to start measuring that. But

26:11[sighs] >> yeah, I don't because it's not recurring. That's like the biggest thing is like we all love this like >> X amount per month. I make 10K a month. Blah blah blah blah. You know, like that's the words that we like is the per month. And so this is like it kind of goes away. Um the first person I ever like like heard this idea of like outcome based right was the guru online gurus who are like you pay when we you know when you get a result or the cold emails that like you don't pay for anything until you get a booked appointment you know it's just like all this soup of um not SAS but service outcome based pricing um is that like who popularized this and I'm not talking like usage necessarily I mean like actual like

26:56outcomebased. Why is it like that the new fad? Does that make sense? Like it seems like a new >> way of doing things and it and it's more trendy if that makes sense. >> Yeah. I mean I couldn't tell you like the exact origin. It's sort of like where did Coronairus come from? >> I want the moment. [laughter] >> Um yeah, we don't know the exact moment, >> right? Yeah. I guess it did take a while, but we did figure that out. Um, no, I think that that's again probably just a uh that's like a tactic and something that's like, oh, this is the way that everything's going and people are getting tired of just paying for retainers that didn't go anywhere. And so they're like, oh, if I can guarantee you this, which I think

27:41there's a lot of merit in that. But at the same time, um, yeah, that was probably something that was just like popularized by some guru like you mentioned. Uh, a similar thing I think has happened in SAS too where some people are like, "Hey, outcomebased pricing is the new big thing. Look, look at this." And I'm not going to name names, but I've seen some people on LinkedIn be like talking about it as if it's already set in stone, but their examples they give are like a there's like a few to several and they're all super wellestablished companies that made a really big bet and we still don't know how it's going to play out, but like 95% of all startups still today are still on the traditional SAS model. Um,

28:25so I think it's kind of silly to be like this is the future. This is inevitable. This is 100% what everyone's going to start doing because we haven't actually even seen that play out. >> Okay. >> Um, go ahead. >> Go ahead, Zach. No, you go. >> No, you. No, you. >> Fine. Fine. Fine. Um, well, how does this how does this play into things like the way that you market your SAS, the way that you sell your SAS? If you uh you know we hear a lot of talk about things like PLG versus SLG and it seems like this has some trickle down effect to those areas. What are you seeing as far as you know is that line continually being blurred?

29:13Are they merging together? Is there maybe it's not even like PLG versus SLG anymore? What's the trajectory there? Yeah, I think that the AI usage is also further popularizing the PLG model because people want to get in there and try for themselves, especially for this kind of AI uh usage and AI based pricing that everyone's going to have to tack on to their existing models where you can't just sell AI over a phone call. You have to get people to use it, but at the same time >> based on, you know, the amount that they're using it and how they're using it, the requirements for using it. We get to things like HIPPA and SLAs's and data and is it is the AI training on my data or not and how is it containerized

30:04and all this kind of stuff that still also warrants a sales conversation and sales process essentially. And that's why I think it there's kind of not going to be PLG or SLG anymore. Everyone's going to have to adopt both because everyone really needs to get people in the door to try and use their AI in order to probably even just get them interested in the first place. And then they're also going to inevitably [clears throat] need to have a conversation around all their requirements for the bigger companies. Um, so I I really like I've always been a big believer in PLG. I've never thought that it was going to like the whole world was going to be completely PLG. And I think AI is going to make this really really clear where

30:51now the future is is everything is going to be both. Why would you ever kind of um handicap yourself for put a ceiling or a floor on who you can attract and what prices you can get, what the types of customers you can get when people are going to expect that both of those options are going to be available. Like 10 years ago maybe we saw a lot of companies take the strategy of like no we don't want to have a enterprise sales process. We don't want to have big customers. We don't want to charge people a ton of money. With AI, though, I just don't see a world where that's ever the case. Like, people are going to be reaching their ceiling all the time. You're going to have to have those conversations at some point.

31:40Yeah. And that makes a lot of sense because you I mean I I think of some a company like Superhum, you know, where they onboard every single customer regardless of their size and that with all of these tools now becoming I feel like the tools are becoming less expensive especially and like the manpower behind them is becoming less expensive even if the profits aren't there or showing it from the >> the actual like uh financial side of things within the SAS company itself, but as far as getting people to use your product and making it more widespread, everybody wants everything. It's like what are you going to compete on? Your customer acquisition cost is not going down guaranteed. Like ads are going to

32:25cost more. Sending emails is going to cost more. You're going to have to do a lot more to send out in the crowd and get people to at least get your foot in the door. And once you do get your foot in the door, why would you stop yourself from giving them the opportunity to hop on a call and get onboarded and, you know, get on get into a salesled motion with that? Even if you like only have to do that and then it's off to the races for them. They no longer have to speak with the salesperson or even sign a contract. They just need more information up front before making the buying decision. >> Yeah. Do you think that this would apply to companies who are purely enterprise level like you know I'm thinking um like a San Diego

33:09based unicorn like Kyrie you know like what what does some or someone who only has enterprise deals with annual contracts only how does it work for for them >> or is it is that a totally different conversation? No, I I think it totally is. And that was going to be something I was going to bring up was I think the biggest change is going to be in kind of ushering in a lot more salesled companies to become more productled because again, you can't just sell AI over the phone. People need to see it, experience it, use it. So maybe it's not going to be completely productled per se where you're just like anonymously seeing usage. You're like, "Oh, that's cool. This big company signed up to start using me from for my

33:55product." But maybe it's more a um concierge pilot for example or maybe you're you can sign up have a playground you have a limited number of um >> kind of AI usage and they can simulate the type of thing that you're looking to do and then maybe that prompts them to book a call to start a sales cycle to uh explore more you know personalized experiences using their own data It's it's not a question of you know like a company like Kyrie or any really enterprisey software um they have to let people use their product before signing the contract.

34:45>> Mhm. >> 5 years ago that was not true. That was not the case. Today it is like they have to use the software. They have to be able to >> play with it, get their hands on it, see it, >> and prove that it's going to work for them. And so maybe it's not productled per se, but it's not just going to be let's hop on a sales call for five weeks, five different calls, sign a contract, and then use the product anymore, >> right? They got to be able to play with it, I think. So I thought of less annoying CRM, who I think has a really nice solution to this. They have a fully interactive demo where like and you can access it straight from their website, you know, it's like click in, you can play. I don't know if it has any of the

35:27AI features that they have put out, but we use that all the time, the design team, because we were looking at like actually using their feature features to see how they worked and jump in there and play with it. >> And it was really interesting because then, you know, like my mom is a sales rep for uh pet food suppliers is like, "Mom, you should check this out. Like it's it's really cool." and and you could play with it for free and then if you want to actually use it um just let them know that Zach sent you because I'm a good good contractor and send [laughter] send clients business when I can. Um but I think that that's having some kind of dummy playground. I don't see why anybody wouldn't do that, you

36:05know, like why why not? You know, it doesn't it doesn't make any sense to keep like what are you trying to hide? like you know the way the sausage is made or you know like there because there's nothing that they're eventually going to uncover it. You might as well just let them play with it and then see if they really want it. Mhm. Mhm. Yeah. I think that's the new uh the new paradigm and as it becomes more and more competitive because people are building faster, offering more for less um the waters are pretty frothy right now so to speak. And so, can you really afford to only be PLG or can you really afford to only be salesled? I don't know. I don't think so. I think that that's probably going to be a big

36:48mistake. And everyone's kind of everyone's kind of already realizing like, hm, it's it'd be nice to have best of both worlds. I would like to have the lower end of the market and the top end of the market. And there's no reason why I can't do both. And in fact, a lot of times that can really help with your acquisition strategy. Um because you get a a larger mix of customers. You're not limiting yourself to, oh, my customer acquisition costs are $300, but um I'm only our top end of our available plans for people to sign up for are $50 per month, for example. And now we're like, okay, the math ain't math in super well. You mix in a couple of thousand per month customers and now you you can kind

37:32of recoup your cost a lot quicker. >> Okay. I I have something that's a little bit adjacent to all this, which is not necessarily tied to AI. Um, we've talked mostly about like tactics and strategies and things that we're seeing from like a broader perspective, but I kind of want to give people like something to chew on for like them to think about uh if they are a founder or thinking about how they're pricing their SAS because there's a whole another side to pricing which is like the psychology of pricing and the like the the position and the product the the fit within your competitors. for pricing. There's all sorts of other things. So, can you give me like what percentage is the psychology and what percentage is just

38:19the math and the tactic? Like if someone's like, "Oh, I could think about all these things and like game it so that it's perfect, but they have disregarded the let's say 90% that they need for the psychological purchasing decision." Like how much should someone put into the psychology versus the math? H man, that's a tough one because you know what? I've struggled with this with software for a long time because software is inherently utilitarian. Like people don't buy software for fun. They buy it to solve a very specific problem for them. Um and as soon as it stops solving that problem, then they're out. They look elsewhere. They cancel. Uh it's game over, right? And so psychology I think only takes you so far for SAS

39:08pricing because at the end of the day people are looking for does this add value to me or not? And there is kind of always an ROI equation going on through their head where they're like hm what are the prices? What am I paying for? And what is the value that I'm paying that I'm getting? Um, and so the psychology is like, oh, can he use decoys, the the 99 cents versus the whole dollar, the um kind of like mix of options? It's like, well, I don't know. Like, I think most software companies should probably have somewhere between one and three plans. And then now with AI, there's probably going to be some kind of um add-on. Most of them probably should have some kind of free trial or premium experiences and then they should all

40:03have some kind of custom enterprisebased pricing as well. >> That's like 95% of the strategy. Can I because we we talked on this briefly before and it's something that I frequently think about in the shower when I'm lying awake at night. Is there is there any kind of status that's brought in with software or tools like this? because it does happen in other areas where there's tools, you know, like, you know, DeWalt tools are more expensive than Ryobi or Cobalt, right? Um, so it does happen and I think it you can see it with other areas like even like you know a water bottle which is more of a tool but you know like the simple modern ones do not cost as much as Yeti and it's like what are you

40:49paying for other than the label right? You know, me and Nick right here with the examples. How much was it, Nick? Like $10. >> I don't know. My wife bought it. I have no idea. [laughter] >> It's less expensive, but that's okay. You know, I'm not not splitting hairs. Just I just know that >> I have a DeWalt over on the other side of the room. I'll have you now. >> Well, and I have and then I have Ryobi, so it's like, you know, why I bought a more expensive water bottle and Nick bought a more expensive power drill than than I did. Um, and I was thinking I think that the only thing that works here as far as like a status symbol is if you solve a really big problem that not many other competitors can solve.

41:32You know, like if you are someone who like deals in like an either very secure or highly not volatile um but like precious or just highly valuable area like if you're dealing with uh companies that only do you know $10 million a year AR or or higher um and you have a solution that's for them. You could have in that scenario, you could have the Goliath type of customer which only has enterprise plans. They're all special contract value. Um each one is treated like a custom agreement with uh you know scope of work and service level agreement and things like that. And you could have somebody who is trying to enter in that same space and maybe compete on a more like a typical PLG route and show like no no our solution

42:28is way more cost effective. It's way more um affordable than these other guys. We can do all the exact same things. And I wonder if they wouldn't get picked simply because they don't have the kind of clout that um somebody would be looking for, >> you know, like does that >> register or >> I mean it is util it's util it's utilitarian but it's more from a perspective of >> this other company has a lot more trust and reput reputability compared to this one which is a like a fledgling startup compared to an established player. there. Regardless of what that means, it could be something as simple as look. Um, it could be the way that they price like, you know, we only do contracts versus >> you could sign up for free right now and

43:15try it out. Mhm. Here's where my mind goes with that. It's a little bit um a non-answer. It's kind of a political answer, but um the way that my brain goes with the idea of like brand power in the pricing equation is um not as much about status as much as it is about taste. Um here's what I mean is I think about the way that people think about and compare Chachi PT versus Claude. for example, Claude in their branding and even down to the way that they fine-tune their models and hard code in if you want to call it that certain types of behaviors into their AI appeals to a certain type of person with a certain type of outcome that they're looking for in a particular way, let's just say, >> versus Chatty

44:18feels a little bit more like the layman's tool where it is a lot more mass market. It's a lot more broad, generic in that sense. It's less opinionated. Even their branding is like, you know, hyper minimalistic, especially for something like chat where everything is sort of like on a grayscale >> and there really doesn't feel like you're getting anything opinionated from OpenAI. um versus maybe even other models like Gemini or uh let's just say like Grock for example where people know that Grock is a little bit more unhinged and maybe it's willing to do certain types of content or responses that other models wouldn't. Even we have like now there's a lot of models where we're we're kind of seeing

45:06the divergence of models and what they're good at and what they're used for because they keep having to fine-tune for certain use cases that they do inherently become different and maybe there's some pricing power in that from the model perspective too where it's like I could I can envision a world and I've kind of already experienced this where sometimes they'll uh a SAS company will tout like oh we have AI features And then I can see that it's OpenAI 40 Mini or something under the hood, [laughter] which is like, you know, kind of a piece of garbage by now. And then I I look at another tool and it's like, no, we only use top-of-the-line models and or we have our own model that we're fine-tuning. I think about like Cursor,

45:50for example, they made their own model um recently called Composer 1 and it's I believe a fork of a Chinese model. I think maybe even like Quen possibly, but they're fine-tuning and personalizing with their own opinions and taste. >> So, there's maybe something there around >> interesting >> like Nike vers Adidas or >> same for Hermes. >> Yeah, but the Well, the strange thing about that is so Chad GPT I'm looking at their pricing right now and Chad GBT is free to start. So is Claude. Um, Claude's Pro version is $17 per month annually. Hang on, let me see here real quick. >> I think it's 20 per month, I think. So, is chatbt it >> it's Yeah, they're both $20 per month of build monthly. Um, but

46:41>> well, that's totally different separate equation as far as showing your annual price versus your monthly price to start. Claude watching you. Um, but then Claude's max is $100 up. So, it's interesting. There is they are playing a little bit of a game and I'm sure they're aware of like they're making it difficult to decide between their high-end models and the high-end packages, but the baseline ones are exactly the same. But >> bite-size. Yeah. >> Yeah. But bite-sized. >> Wow. Um, okay. We're coming up on time. Is there anything else you wanted to say before I I I finish out with what I wanted to say? >> No. Um, I guess like my my parting words are think about this really thoroughly. Talk

47:25to us if you need some help thinking through this. Um, but but really honestly like this can basically make or break your business if you're a SAS company over the next couple of years because again if you you could unknowingly be eating a lot of the costs and then you're not profitable andor not up to snuff um with your competitors and how they're rolling out AI features or bundling um all their pricing and and packaging. Uh but number two, it's a huge opportunity as well. And so like get this right and it can be a huge growth ever for you. It can make you a lot more profitable. Um and so it's important to like think about this hard now, bro. Look at this guy. Put it on a poster. Um okay, thank you. Um

48:16before we go, we have Nick's dumb idea of the week. Okay. So while we were talking, I had something that hit my head which it uh related to the tokens, buying tokens. Okay, so there there's the story. Have you guys heard like how Sam Bakeman Free like got rich at the very beginning? How he like was buying Bitcoin in like the US and like selling it in China or something and there was like a just a slight difference? Yeah. Japan. And so he made like >> Yeah. arbitrage. Okay. So, where is the AI token arbitrage and can we capitalize? Like, where can we like bulk buy a bunch of tokens and then like sell them somewhere else where there is money? That's my dumb idea of the week. Okay. If you guys have an idea, you

49:02could chew on it. Let me know. >> Not a dumb idea, actually. And it's already there's businesses being made off of this. >> Dang it. >> Yeah. where they basically make themselves uh a gateway or a router for any model that you want to use and then instead of paying directly for each one of those models or providers, you pay them and then they just have a teeny teeny teeny tiny upcharge, right? It's sort of like >> I'm a genius. >> Credit card processing, you just add 3% on top. >> Gosh, >> I knew it. I was like, "Dude, there's got to be some You can bulk buy AI tokens." Like, come on. This is like This is Bitcoin is here. That's exactly what it is to to anyone who doesn't actually know how to do it. Dang it.

49:50Okay. Well, Nick's dumb idea of the week turned out to be a good idea. I'm just like probably two years late. [laughter] >> Um, okay. Uh, well, thanks for, uh, listening to the factory floor and we will see you guys on the next one. Peace out.