Building AI Pricing with Dan Balcauski

But now we're in a new era, and in that transition, the idea is that we have to be agile and understand that out of the gate we're going to face all the problems we always face with introducing new products. Every Chief Product Officer out there knows, or CEO knows, the roadmap from last year is littered with a bunch of features that you had high hopes for and for whatever reason just didn't get the adoption that you expected. Your AI features are going to be no different, but now you also have this underlying cost model plus new competitive pressures. This is SaaS scaled, the podcast where data meets action with host Arman Shrakhi. Each week Arman will be sitting down with CEOs and industry leaders from the technology sector, giving you the insight to innovate without reinforcement. Inventing the wheel they'll discuss challenges, best practices, and how to identify the right metrics. So if you want to get to market faster and in a way that matters, then subscribe and join us every week as we discuss SaaS scale. This episode is brought to you by Curve A, the modern no code analytics solution. The tools you need to take action with your data on a platform built for maximum scalability, security and cost efficiencies. If you're ready to reduce complexity and dramatically lower costs, then contact us today@kurve.com that's Q R V Y.com. Hello, welcome to another episode of SAS Scaled. I have Dan with me and he's a second actually time podcast guest. And it's not because both of us live in Austin. Believe it or not, we haven't yet had a chance to really grab coffee and meet in person. But it's because what he talks about is super attractive and interesting and impactful to SaaS companies and I would love having him on the show and just ask him some questions that would help SaaS companies. Dan, please take it from here. Introduce yourself and then tell us why this topic that you are expert in is so important. Yeah, well good to be back Arman. And we'll have to fix that oversight on my behalf of grabbing coffee in person. I think this is our third conversation since you also came on my podcast. So really happy to be back. As you mentioned, name is Dan Bachowski. I run a consulting firm here in Austin but serve clients pretty much worldwide at this point. Focus on helping CEOs of scale up B2B. SaaS companies define pricing and packaging for new products and I think that intersects directly with the conversation today because any new product under development right now in a SaaS business is going to have AI in it and that has really changed the margin profile at a minimum and potentially even the business model of this SaaS world that we've gotten comfortable with over the last 15, 20 years. I think Salesforce was the first SaaS company, the Granddaddy. They were founded in 99, taking on Siebel Systems, you know, perpetual software and you know, it's hard for folks to imagine now, but the transition from perpetual to SaaS was very rocky. Not everyone made it. There's a few really shining examples of companies that made that transition like Adobe and Microsoft and there's a bunch of also rans that never actually made that transition successfully. And I think, you know, we all got used to attending the SaaS docs and sasters of the world and learning the SaaS playbook and what new metric do we have to worry about this year? And we got comfortable because after 15 years there wasn't really much new at those conferences. Not to say anything bad, I had very excellent experiences at them. But a lot of those playbooks are having to get thrown out the window because of disruption from AI. And now, you know, CEOs are under pressure from their boards. Investors are putting pressure on public traded companies because you know, their terminal values on their discounted cash flow models for their valuations is like, well, we used to look at this as an ongoing annuity, but every time Anthropic or OpenAI releases a new plugin for cybersecurity or legal, we see those companies that are in the space take a 15% stock hit that day of the announcement. And so a lot of executives at these companies are really trying to figure out how do I turn this from a liability into a strategic advantage. And so that's where they pick up the phone. Yeah, that's very important. So I can tell you for sure, I have seen some SaaS companies that really, despite all of the innovations that they have made in the product and everything, but the biggest innovation and the biggest, the most impactful one was actually the pricing innovation. When they changed their pricing model, I could see how fast they started growing. And not because they changed anything in their offering, but because they changed really the way they looked at their packaging and pricing. So definitely this topic is very important. And, and for some of you who have not heard the first episode, I would encourage you to go to saskale podcast website, search for Dan episode and watch it if you want because that one, despite being two years ago, maybe three years ago, it's still, I think it's three years at this point. Yeah, Very still probably relevant and meaningful and good topics discussed there. But now let's talk about the pricing with AI in mind. So that's really what Dan has been focused on helping companies with, to come up with the right pricing strategy they have. So one of the things that Dan mentioned to me before this call was that really AI pricing isn't a problem to solve, it's just, it's about a process to build. What do you mean by that? Yeah, well, I think you know, in these, you know, we might call it a phase transition. You know the perpetual to SaaS was, was one. Right. Similar, you know, using it in terms of like when ice, you know, turns to water. Right. It's like, yeah, they are the same molecule but they look very different and have very different properties. You know, there's a lot of change and that change is happening in real time and very quickly. And so, you know, I'm learning along with everyone else and the market is learning in real time as new capabilities seem to get released. I mean if not daily, it seems like weekly. Right. Every new model that the foundation model companies, whether that's Google or Anthropic or OpenAI releases seems to unlock new capabilities that we could only dream of a year ago. And as well as changes in cost models. I was reading a report not too long ago that OpenAI on their like newest model on a per task basis versus the model they state of the art model they had a year ago was 390 times more cost effective. Right. So that doesn't happen in industries very often. Right. If like I'm selling hammers, you know, it's made of wood and metal. Right, Wood and metal. You know, unless we're going through some raging Argentina level inflation crisis is not changing 390x and so you know, with the change of capabilities, the change in underlying cost structure, the fact that people are bolting on new capabilities that customers have never needed to see, have never seen before or really know how to interact with that all implies the need to stop worrying about getting it right out of the gate and instead shift your mindset to okay, we were in an era, we were in the ice era of SaaS and that was baked and you know, some people had per seats and some people had per gigabyte or and you know, it was a recurring subscription and now we got all the billing systems to be able to do that finally. But now we're in a new era and in that transition the idea is that we have to be Agile and understand that out of the gate we're going to face all the problems we always face with introducing new products. Every chief product officer out there knows, or CEO knows, the roadmap from last year is littered with a bunch of features that you had high hopes for and for whatever reason just didn't get the adoption that you expected. Your AI features are going to be no different. But now you also have this underlying cost model plus new competitive pressures. And so I think really what we need to embrace is the fact that pricing along with the change of value and cost is really going to require an agile mindset and the idea that we're going to ship, we're going to learn and, and a lot of companies became not complacent in a bad way. But I mean there was again, the playbooks were pretty much baked. We knew what the metrics were, we knew what the margin profile of SaaS companies was going to be. And a lot of that, given that's going to change, we're going to have to become more agile of building those systems that allow us to ship iterate change quickly. And with, you know, AI coming to market, it will accelerate some of these developments for SaaS companies. And SaaS companies really can move forward faster with their product innovations. How does it impact the pricing from your perspective, that accelerated path to evolve products faster? So there's two tightly linked questions there. So, you know, the first is, you know, when you add an AI capability to your product that does not equal value, that may very directly translate into a cost line item that your CFO is going to be like, oh my God, look how much money is going out every month to OpenAI or Anthropic. But just because we add AI does not mean we increased value. And there's a, I think big set of lessons that I think the industry is learning in real time in how to translate these capabilities into tangible user value. So that's on the, you know, the surface value, the surface level where, you know, just because I stamped, hey, this is AI enabled, okay? Every company has done that, right? So we've all been AI washed. So your marketing messaging around that is meaningless. And AI is purely a enabler of enabling technology. Right? We don't, you know, as SaaS businesses, we've never gone around and said, well our SaaS company runs on an Oracle database backend or redshift or you know, snowflake, like your customers don't care, that's your enabling technology. I think this was the mistake that a lot of the companies in the web 3 crypto era made was like, well, it's like you're, it's like Twitter, but it's on the blockchain. And everyone's like, I don't get it. Right. So we don't want to do that. Right. At that surface level. But then also, you know, there's a really steep level of adoption. You know, this goes all the way back to, you know, those of us have been in technology a long time, have probably read the classic 90s business book by Jeffrey Moore, Crossing the Chasm. There's different stages of market participants. The early adopters, you give them Claude or even Claude code at this point. They're like, oh my God, this is super powerful. You give a Claude chat interface to my mother and she's like, what do I, what do I do with this? They treat it like Google. Right? And so like you built in a chat support agent into your product or it helps, maybe you sell an analytics product and it helps you build dashboards just by typing what you want. Yeah, your early adopters getting a ton of value, but that's probably not the majority of your revenue. And you need to figure out, okay, how do we bring folks along to actually realize the value we think we're going to attain? And I'll also just say as part of this, you know, when we think about pricing, you only get to price your differentiated value. The market will set the price for undifferentiated value. And so we have to be very careful about when we're introducing new functionality that we're understanding that, okay, our customers don't care about our costs, our cost profile changes, but we just added capability that all of our competitors either have already or will have in the next two quarters. And if we don't all raise our prices and you can't coordinate because that'd be collusion. It's highly illegal in pretty much every jurisdiction around the world. You don't get to price for that value. So you have two structured problems there, which is you have to understand, are your customers actually getting the value that you're promising them because of this adoption risk and learning curve? And then is that value that we're delivering, you know, differentiated? And I'll even tie that to, you know, I've been helping companies with adding sort of copilot like assistants into their products. And in customer research conversations, a lot of companies or prospects will say, well, look, I would never pay more than $20 for this because I could basically get the same thing from ChatGPT. I'll just export my data, I'll give it to ChatGPT and I'll have ChatGPT. And so you're also competing against the power of these fundamental providers. So I think those factors are all increasing the complexity of this situation. I think there are multiple parts to your question, and I think I only touched on one of them. Yeah, but there are two aspects to AI, right? So from a SaaS company, you use it as a productivity tool and that makes you faster to really go forward and improve your product better, faster, sooner bring it to the market. And that's, as you said, behind the scene, customers may not really even see it or care about it. They just see the outcome that is coming to them faster. The other aspect is you add AI as a capability, as a feature, as a product, depending on how you want it to really present it and what your product is, and then that adds value probably to the customer immediately because if you design it correctly, probably you add it because there was a value there. Right. So that's, that's two aspects. Now my question was more about really the kind of techniques that software companies now have access to. Right. You can create agents to do your QA better faster. You can maybe get some code done. This coding can be done by agents in a faster way. You can really review your code from security perspective faster than before because of these AI technologies, all of them that are internally being utilized now by SaaS companies. And that accelerates their process, that accelerates them in a way that they can bring the product to market faster and build better products sooner. Does that kind of, how does it impact the pricing competition between different SaaS companies with competing product? Do you believe that that is something that would make any difference or it's just, it would be the same as before. The same formula that you had in the past from competition perspective will apply still. The case is not going to change. It's just going to be at a faster pace. Yeah, it's an interesting problem. I think there's definitely a narrative right now on Wall street that this is going to significantly impact these companies because we're seeing valuations just get hammered on what seems like a daily basis of soft, the basket of public SaaS companies. And I think there's a couple of different narratives at play. So one is the foundation model. Companies, whether it's anthropic or OpenAI, are basically going to create AGI or near to it, and therefore nobody will need anything except your own, you know, anthropic account or Claude code. And just you could you know, just like asking the GENIE for your wish. Just say, hey, I need a CRM today. Can you build me one? I think that's a bit overstated. I think that really shrinks the idea of what a SaaS company is in this day and age. And maybe I'll use an analogy from the transition of perpetual to subscription. So perpetual, we used to sell, you know, a license and you'd buy the software. When we moved to SaaS, it was subscription, but it was also hosted. And there's a key word in SaaS, software as a service. So a lot of companies that didn't make the jump between those two didn't realize that there was now a serv. They were not software companies, they were service companies. What did that do? It created roles we're now all familiar with, but maybe forgot that we created along the way. Customer success. I need to make sure, you know, I used to be able to sell an enterprise subscription platform. I got 90% of the customer lifetime value on transaction. I don't care if you ever actually install it. Like, that's not my problem. Like, yeah, I got my money, good luck. But now we actually needed to make sure customers not only used it, but drove the expected business results because the subscription model required them to renew. We also had things like DevOps. DevOps didn't usually exist. Why? Because now we can't just, hey, here's the software. Tell your IT people to go provision a server, install the software, run the updates, make sure if there's downtime, they have someone on call to fix it. Now the software company embedded all that learning, training, all of these things became part of subscription. So I think if we just say, hey, everyone's going to vibe code their own salesforce. And for, it's like I think really shrinks the importance of all of the other functions of a SaaS company that is not just writing the software code. Right? And so, you know, I think that is a narrative outside. But I think you, you know, back to your question, you know, will we see more competitive pressure as software companies can shift ship features capabilities faster? Maybe I haven't seen it yet. As much as everyone's talking about the amazing coding tools, whether that's Codex or Claude Code or Cursor or Windsurf, I don't know about you. I am a very avid watcher of the software market. I have not been blown away by all the software that I'm seeing companies ship. So I think there's a lot of quick to prototype. I have not seen a full Unleashing of the bottleneck to full production software, I think as well, it's getting more and more important on discernment of what should we build. Everybody knows the situation. You know, if you're a SaaS CEO or executive listening to this podcast, you build out some amazing capability to satisfy a customer's job, jobs to be done. And then six months later your competitor says, oh, we have it too. But then you go look at what they actually built and you're like, yeah, they built like 10% of that. That actually doesn't get the job done, but it allows them on their marketing site to say, oh, we do that too, because they were, you know, losing deals because of it. So I think you end up in a situation where it becomes easy for maybe an engineer to go, oh yeah, like let me go fill that gap. But that's also, again, not understanding what real product management and software development and design is about. It's about really deeply understanding that customer's context, their problem, the job they're trying to get done, their key outcomes, and building an entire system that helps solve that for them end to end. Not just oh, did we do enough so that we could add that check checkbox to our our pricing feature page. So I think we've got a ways to go. I think that will happen over time. I'm not seeing it yet. So my sense is that's probably a five year now problem versus a right now problem. So when you look at SaaS companies and you have worked with plenty of them to advise them on pricing strategies, I bet there are some pricing strategies that are quick to address some of the problems that they have, but then they may not be happy about it in the long run. And there are some other pricing strategies that may not be super easy to implement at the beginning, but it pays off in long run. Can you give us any example of the pricing mistakes that you have seen people make? That is quick at the beginning, but in 24 months or so the SaaS CEO comes back and said, that was not a good decision, it was easy, I made it that way. But now I understand it was not a good decision to make. Yeah, well, I think that's a great question and I think it ties back to what we were saying earlier on in the conversation, which is really important to have this mindset. So I've done a bunch of research on companies releasing both copilot assistance as well as agentic capabilities. 100% of them have revised their pricing approach within 18 months. Salesforce, I think, is the Most extreme example, they released their Agent Force platform a year and a half ago and in 18 months they've done three major pricing and packaging changes. So this goes back to the mindset issue is like you should expect that you're going to get it wrong out of the gate. 100% of companies who have already gone down this path have already revised within 18 months. So plan for that. Don't think you're going to be the exception. I wish you the best of luck. I hope you are the exception, but don't plan for that. I think the number one mistake that's super quick is pay. We're about to pay Google, Gemini or OpenAI ChatGPT for all of these tokens. They're billing us on X dollars per input, Y dollars per output token. We are going to do a cost, either just a direct cost pass through or a cost plus pass through to our customers. And this usually looks something like I'll use a CRM example since everyone is familiar with CRMs for the most part. Hey, we price per salesperson $10 a seat for our basic plan and then we have email summarization and email generation capabilities and each seat gets 1,000 tokens per month. Looks really simple. The problem is it's pouring sand in the gas tank of your go to market engine because now every salesperson has to one, get educated on a token and they now have to be able explain that in a way that customers or prospects can understand. And there's a general rule about pricing metrics which is we generally want to make them one easy to understand and ideally something that those prospects already count. The reason why seats really work well in a CRM market, there's a couple but primarily all of those reps are quota carrying so it ties directly to the size of the revenue of the company. And two, I know how many salespeople I have if I'm talking to a CRO trying to sell a CRM, be like how many salespeople you have? I've got a hundred. That's $10 a seat, $1,000 a month. Sweet. If you have to be like okay, now how many tokens are they going to use? I have no idea. So we call this showing your customers your underpants, your tokens that you're paying OpenAI are your costs your problem? Your customers want to know about them. They don't care, they don't want to be educated about it. And even we have some really, really brilliant B2B salespeople out there. Even the best of them if they could explain it and get the heads nodding on the other side. As soon as that prospect hangs up the phone, 10 minutes later, they're going to be busy with something else. And then a day later, when they come back to it, they'll be like, I don't remember anything that guy said. And you do not want to create those points of confusion with your customers. So then I can go on. But then the question is, what do you do instead? But I'll see where you want to go. Sure. The way we see AI in many cases is that it's not really competing with software. Actually, AI needs software and uses software. I would argue that with AI agents in place and more digital workers in place, probably the usage of software would explode in the coming years. Meaning that rather than just 200 users go to the CRM now, there will be 2 million AI agents also hitting on the API and using the CRM software to do something with it, to get the information, to book something, to add something, to log something, to analyze something, whatever information exists there would be also utilized by these AI agents and sub agents. Now the question is that if the pricing today has been designed good or bad in most cases, per seat, and for really humans now, what happens when these AI agents are going to be in place? Increasingly every month, every quarter, we have seen more digital users come to the place and they use the software at different pace and different style, they are faster, they work 24 by 7. They, you know, 24, 7. They really kind of, you know, utilize the software, you know, at a totally different level. So how do you think that kind of, you know, pricing will shape up in the, in the, in the market for SaaS companies? Yeah, there's a, there's a bunch of interrelated thoughts there. So I think one way to look at it is I would separate maybe what we were referencing earlier, which is like AI co pilots. So this would be your human interacting with a chat interface like ChatGPT. I ask a question, it gives me an answer. We go back and forth, we iterate, brainstorm. I would separate that from AI agents, which was the sort of foundation of your question. And I think you outlined it really well, which is we have to understand that these agents are not beholden to the, you know, they don't sleep. It's a very much a terminator. They don't sleep, they don't stop, they feel no remorse. Right? They will just churn and churn and churn. And those then have different, very different usage profiles. And so we want to make sure we separate those cleanly when it comes to the idea. I think I've seen one really nice pattern play out which is, you know, when we're adding these capabilities, are we enhancing value driver for an existing customer or use case, are we adding a net new value driver? And you know, go back to the CRM example. If I'm adding an ability for hey, I have, I've got an account, you know, we've had this customer for two years. There's a hundred emails from five different sales reps and now I've got to go in and try to sell them some more software. I don't want to go dig through all the emails or PDFs attached to this account, different quotes they saw. I want an AI to come in and summarize all that so I can get up to speed. Right? That capability is enhancing the value driver that you already were selling that seat. So we're making that salesperson, that rep more effective, more efficient and, and therefore it's should generally already be aligned with the pricing model in that CRM per user. Now if it's really effective we may do things like considering, okay, we've added this capability, do we increase the amount of money we're charging per user? Because now we've supercharged that person, right? What they could, what used to take them a day to do, they can now do in an hour or maybe 10 minutes. And so they're getting more value through using our system and therefore, you know, we're still aligned by that person to person, but maybe we could charge more. The other option folks are taking with those type of capabilities is we use those as a upsell capability in our higher price tiers. So we've added these AI capabilities to make these agents or salespeople more effective. We put that in our gold tier. So you need to upgrade to that, right? And so we'll see an increase in revenue that aligns with our increase in costs but our customers will get the appropriate value. So I'll separate that from is this a net new value driver? And I think, you know, one of the best examples of this is intercom. So on there it's support, not sales, but similar idea where they release their fin AI agent there the value proposition is we're allowing this software agent to deflect tickets from our humans. So there they're not trying to tie it to the per user model but instead they're creating a separate dimension because it's a brand new value driver and so it's separate even on the packaging side. From, you know, it's not included in sort of their good, better, best and there they chose a 99 cent per resolved ticket output. Now I think where folks are maybe getting tripped up and this ties back to our previous question as well is that there's a, a general, I would say confusion in the market because people talk about Intercom and Fin AI as outcome based pricing. Outcome based pricing is not new. It's been around for a very, very long time. Your, your sales, your, your real, your realtor uses outcome based pricing. They get a commission on the sale, right? They get a percent of the sales price. That's outcome based pricing. It's been around for a long time and it's, it's predominant in a lot of markets but it's very hard to pull off in practice. The outcome the customer cares about for B2B usually boils down to save you money, make you money for customer support. That VP of customer support, Chief Customer Officer cost of support is usually a line item and they're saying you don't get any more budget, you're going to have to handle 10x more tickets this year. And so they're really trying to increase that efficiency of their staff or maybe even make do with even less money because it's treated as an expense from the perspective of the business. But Intercom, Finai, they can't promise you we're going to save you $10 million in customer support costs. We can only track to an output. And so I say my current mission in life is to try to help educate the marketplace on understanding that what they're doing there is not outcome based pricing, but I think more appropriately output based pricing. And so we think about the world we are in and a per user model, the users still need to take all the action to ultimately drive the outcome, including all the intermediate outputs. Now with these agents we're able to move farther along the value chain to create these intermediate outputs. Ultimately we'd like to be able to create outcomes, but there's a whole bunch of complications I'll leave to the side. And that is how you get out of that token trap that was handicapped before. Right? Because tokens are a pure input metric that again, your customers don't care about. So the companies that are really doing this well have thought very deeply about what are these intermediate outputs that those customers are already measuring and tying it back to something I said before. Why does that fin AI $0.99 per resolve ticket make a lot of sense? Because that VP of customer support probably knows what their daily Ticket volume is they already count that. Oh, you say we'll charge you 99 cents per deflect ticket. We've got a 88%. Those requests that come through the web, that guy's probably already tracked or gals are probably already tracking those metrics. He goes, oh my God. And he's doing the math in their heads of what that would mean to them. So. But you say, hey, you know, when our agent runs, it costs you 300 tokens for a complex query, 200 tokens for a simple query, and all of a sudden you're in no man's land of needing to explain your product architecture and OpenAI's back end and they you've just totally lost the sale. Great point. So I'm going to end our discussion with a fun, unrelated random question. And that would be amongst all of the cities that you have traveled and you have been there, which one you enjoyed most and you would like to, you know, if I ask you, hey Dan, what was the city you enjoyed most so I can go there and enjoy, you know, the time for two weeks? You would say, oh definitely. I would, I would tell you to go to this city because it's amazing. Can you share with us your favorite city that you liked and enjoyed? Yeah, I would say Buenos Aires, Argentina, probably my number one. Followed closely by Barcelona, Spain, but I would say Buenos Aires, number one. I've had the chance to go a couple times and I did. I lived there for three months at one point a few years ago. Nice. Thank you very much for sharing that. And again, thanks for being with us as always. Great speaking with you. Likewise. Thank you for having me. Thank you for listening to SAS Scaled with Arman Ashragi. For show notes and any resources mentioned in today's episode, go to SAS scaled.com if you're enjoying our show, give us a five star review and share on LinkedIn. And be sure to subscribe for any updates on future episodes. Thanks for listening. This episode is brought to you by Curve A the modern no code analytics solution. The tools you need to take action with your data on a platform built for maximum scalability, security and cost efficiencies. If you're ready to reduce complexity and dramatically lower costs us then contact us today at curve. Com, that's Q R V E Y Com.