Pricing Page unPacked - NOTION: AI credits and the move to value-based pricing

Rob Litterst (0:17)

Welcome to Pricing Page unPacked. I'm Rob Litterst and I'm joined by Ulrik Lehrskov-Schmidt, pricing expert and CEO at Willingness to Pay. Each week we take a real company's pricing page and break it down. The decisions behind it, the trade-offs, and what it tells us about how the company actually wants to grow. No slides, no scripts, just a real conversation between two friends who live and breathe pricing.

Ulrik Lehrskov-Schmidt (0:38)

I'm excited for this one, Ulrik.

Rob Litterst (0:40)

I want to talk about Notion because they just made a really big pricing change. And I understand that the pricing, you were visited in your dreams about Notion pricing recently.

Ulrik Lehrskov-Schmidt (0:50)

Yeah, I literally dreamt about it the other night. So I was like, okay, I think we're going to cover Anthropic or one of these, but then it was like, okay, like when you dream about something. And it's one of these where I think we had a podcast about Notion a few months back. And I actually haven't revisited that. But from memory, I think the conclusion was something along the lines of, okay, super cheap, super like basic stuff. They're kind of at a crossroads where they need to like decide if they want to just like keep expanding their customer base with like cheap flat-ish pricing, or if they want to, and then sell to someone, or if they're going to turn on monetization. And so I think what they did was they started to turn on monetization. So it's kind of like a strategic choice here. So really excited to sort of unpack how they've done that. It's a really interesting change.

Rob Litterst (1:43)

I think like Notion has this weird balance that they have to straddle where they have a lot of users who are not super technical, right? They're like using Notion as kind of like a productivity app or like more or less like a note-taking tool with database functionality and stuff like that. But then they have other people who are using a lot of their AI tools. And if you look at what Notion is trying to become in the AI era, it is a lot more than a note-taking app, right? So like there's a very different value prop that they're really going for here. I think that's where a lot of this tension comes from. I figured, so I wrote about this change last week, Ulrik, and I figured what we could do to kind of set the stage. The newsletter write-up that I did, I broke down kind of like how they've been approaching AI pricing. And I did like kind of a historical timeline just to recap that I think would be helpful to setting the stage. And then I can try to explain to you what they're doing now. And I would love to get your take and we can dive into the pricing page and all that good stuff. Let's do it.

Ulrik Lehrskov-Schmidt (2:39)

What kind of like sparked this post last week is I got an email because I'm a Notion customer.

Rob Litterst (2:43)

So I want to make it very clear and like be very transparent with listeners. I use Notion a lot. It's one of the few apps that I use every day. I'm in there all the time. And I love Notion. I've been using it for a long time. It is a great app, and I'm a very big fan. So I want to get that fanhood out of the way first, because I'm about to be kind of critical of their new pricing. But it comes from a place of love. So they sent me this email last week breaking down their new pricing change, and my first thought was that they are making their new pricing model unnecessarily complicated. Here in a moment. Before we get there, I want to just give a quick primer on where they've been from an AI standpoint. When Notion first launched AI, they launched it as an add-on that went alongside all of their other plans, and it was kind of like a flat price add-on. You pay $8 per member per month if you're on an annual plan, and you pay $10 per member per month if you're on a monthly plan, and you get all this access to Notion's AI features and functionality, which at that point was really more or less like chatbot Q&A, auto-filling, and like writing assistance. So this was kind of like in the era where like all the legacy SaaS apps were launching AI add-ons, and Notion was one of the first to get that out there, and it seems like it was pretty successful. The second iteration of Notion's AI, which is when they bundled it into their business and pro plans. So they got rid of the add-on, they bundled it into kind of like their top tiers for business, and they offered a trial in their lower tiers, free and plus. And so you can see here, Notion AI is, it includes their agent, enterprise search, and AI meeting notes. And I want to mention like right around this time, I was talking to a few friends who run SaaS businesses, and they literally said that the enterprise search in Notion is like the most important functionality of like any tool that they use, because their support team uses it constantly for customer support. It's like their repository for any question, period. And so they were using Notion AI quite a bit to help with customer success and customer support. Okay, so we go add-on, bundle, and then we get another add-on here. So after kind of bundling core AI functionality into the pro and business plans, Notion launched a custom agents add-on. And this is really where they introduced their credit model. Notion has an in-house agent that comes with every business and enterprise plan that will do like all sorts of kind of like managerial and like menial tasks for you. Custom agents you can build yourself to handle repetitive tasks and to do work that you're doing constantly, and they charge you for that for credits and so you can totally customize these agents and kind of like add them to your Notion instance and then you pay $10 per one thousand credits and so this is that's when they introduced this. Okay and then we get to this latest change. So before this change, they didn't have a visible credit balance for Notion Agent, right? So Notion Agent just kind of worked in your plan, did all of these things for you, and there wasn't really like a visible or tangible credit balance for what that agent is doing. With this introduction and announcement that they made for their latest pricing, they say that most people will not hit their limit for Notion Agent. But that's like one of the big things here. So like with this introduction, they now have two different credit types. So they have standard credits and select credits. I'll actually go to the pricing page right now to kind of hover and show you what they mean. For those of you that are listening and can't see the screen, I'm hovering on the Notion pricing page. So standard credits are allocated per seat and they're used for work where Notion picks the model. Like Notion agent on auto being their AI router and their model router. We'll talk about that in a sec. Meeting notes and database autofill. Select credits are also allocated per seat, and they're used for work where you pick the model, plus custom agents that you own. Lastly, there is this concept of Notion credits that admins can purchase to cover additional AI usage, and Notion credits can ostensibly be used for either standard credits or select credits. I know I just went through a kind of a long history of Notion's AI pricing. What is your first thought when it comes to this new update, I guess?

Ulrik Lehrskov-Schmidt (7:13)

So a few things. So what's the timeline? Like, is it a year and a half since they're like going through all these different stages that you just narrated? So this is Q1 2024.

Rob Litterst (7:23)

They expanded on their AI add-on. So I think that even means they had their AI add-on back in 2023. They bundled AI features into their paid plans in Q3 2025. So they had an AI add-on for a long time. They had it until like halfway into 2025. That's when they launched their, that's when they like bundled it in and really kind of made it like a core part of the platform. And then they launched their custom agents add-on and kind of kicked off their credit model in March of 2026. So March of this year. So they kicked off credits earlier this year.

Ulrik Lehrskov-Schmidt (8:01)

Yeah, again, for those of you listening, so we're looking at the time series that we have on PricingSaaS where we capture all the pricing pages every And day. So Notion is logged with like three pricing changes in 2024, three pricing changes in 2025, and then 14 year-to-date in 2026. So it's like this acceleration of pricing changes. So when you said, what did I notice from everything was you're like a customer and a fan and you're not, let's say, pissed off by any of the changes. So it's like, yeah, we have a company, they are navigating AI and they're doing a lot of different things. And in a way, it's not like deteriorating the relationship with you. So it's just like an observation where a lot of people I think are either implicitly or explicitly afraid that change is bad. And then if do change, we kind of chip away at the relationship with the customer. And I think there's this moment in history right now where there's a window of opportunity for pricing change because it is so externally driven. So if I come out right now and say, hey, we need to change pricing because AI, everybody is going to say, yeah, sure. Like, that's obviously not your fault. Like, it's like you're just having to react to this. So it's a little bit like you're not blaming the gas station on the corner if gas prices go up because oil prices go up. It's like it's not, you know, it's not Joe's fault. Like he's just running the gas station. So I think it's a little bit here again when Notion is raising prices or doing something, even though they might be very proactively engaged in finding a model that works for them. And I think what we're going to see here is that Notion is going to find a little bit a similar version of value unlock that Figma did. So very traditional like seat-based flat pricing doesn't monetize a lot. I think we went over the unit economics in the last podcast where it's like, okay, X number of users have a lot of users multiplied by a relatively low revenue per user gives whatever their ARR is. And they are likely going to have parts of their customer base that are willing to pay a lot more than they are now. They're so integrated. They're so, they're so, Notion is so useful to them in the way they do work. But they only have the seat-based model, just as Figma just have the seat-based model. So if we suddenly, through agents, through AI, find a way to monetize the consumption or the usage or how many times a table is updated or does something or like data goes from A to B to C, Suddenly, we're going to see some customers paying, you know, 10, 20, 100, 500x of what they did before and still being sort of paying according to value, not wanting to churn. But the price per seat-based model was never able to get at that value. So I think what we're seeing here is that Notion, or they're kind of getting like a value-based pricing in through the backdoor, through the narrative of AI. I'm not sure that they're doing it on purpose. I think they're partly doing it, as I think Figma was, forced by cost. Like we have to give AI to remain relevant. AI costs money, so we can't give it a flat rate. Hence, we have to price for AI. Oh, that model turned out to also be really good at like actually monetizing the inherent value in our product. Yeah. So I think that's a little bit the story that they're going through here. Then on the actual execution level, with these three types of credits in hierarchy, like standard credits, select credits, and Notion credits, I think they're still a little bit stuck in something that I think they will regret fast, which is that the credits are tied to seats. So if I have 10 users, I get 1,500 standard credits per user, so 15,000. But as I read their model, you know, Mike can run out of credits when Sarah still has credits left over. Yeah. So they're inherently like non-fungible across seats. And then I'm seeing these Notion credits come in as this sort of account level pool that can then sort of fill up each of the seats. And that just seems to add too much friction. Like if I have like, so because what's going to happen is that if I have a hundred Notion users or a thousand, whatever it is, I'm likely going to find that a lot of these Notion users are not using their credits. Half of my users are not going to use the standard credits at all. For them, Notion is a place where they go and look at something whenever their co-workers tell them to. It's like, okay, I have the data, I'm going to go back to whatever it is I'm doing. So if I then have, but some of my users are going to be hyper users, they're going to use way more than the 1500 credits. So now as a customer, I'm forced to go in and buy additional credits, even though I have credits. Right. So it creates this unnecessary friction where people sort of hold off or hold off or argue they try to reduce the consumption of the hyper-users, which are the ones that actually drive adoption of the product and create stickiness and create value and all these things. If I look at it a little bit, I think they're sort of on the evolution path towards something that's good, but they're not there yet. I think you're right.

Rob Litterst (13:44)

And I think what they want to be, I think there's this kind of common trope or kind of common tweet that happens. Anytime Notion launches some new AI functionality, somebody will tweet at them like, this is a note-taking app. Basically, it'll be like, we just launched our agents and workers and all this. Like Notion has all sorts of like crazy AI functionality stuff now like they've really kind of built the product out into a very new direction with AI. But every time they launch something new like that, somebody basically tweets a version of that where it's like, this is a note-taking app. And it's simply not anymore. And I think some people still use it as a note-taking app. But I think where Notion is kind of taking the product is they want to be this all-in-one kind of like both. And this is where it gets interesting, is they want to be both a system of record, which is what I primarily use Notion for, is kind of like a system of record. And they also want to be a system of action. So they want to have this place where you store everything, but then they also want to be the place where you're kind of building agents to actually do the work for you, right? That's a very interesting proposition. And I think a credit model works pretty well for like just very broadly speaking for a solution like that. But I think you'll find like most companies that are kind of working as systems of action have more flexible credit models, right? And you're able to kind of like move credits around more easily. I actually think it's really funny that their kind of like money quote on their pricing page is from OpenAI. There's power in a single platform where you can do all your work out of. Notion is that single place because I do think OpenAI is kind of also trying to be that single place in a way because they've got their new agents well. And I think that's going to be the challenge for Notion's kind of like competing with some of those products, but we can get to that later.

Ulrik Lehrskov-Schmidt (15:33)

For sure. I think for a while, I think beginning of 2026, when I think AI got really good about a year ago, like end of 2025.

Rob Litterst (15:43)

We're going to say that again next right?

Ulrik Lehrskov-Schmidt (15:45)

A year ago. When it got good. But anyways, like right in our current timeline, you know, end of 25, that's when a lot of like the SaaS apocalypse started and you had like, oh, like regular like SaaS is dead. Valuations started to sort of really like go south. Overall market multiples of like B2B SaaS companies traded publicly went down to not only the lowest point ever, I think it hit like 2.9x revenue, which is like, I think the point like for the last 15 years before that was like closer to the Forex.

Rob Litterst (16:22)

So suddenly it was like, it was way lower than before.

Ulrik Lehrskov-Schmidt (16:27)

And a lot of this was, oh, there are no modes anymore. Even though the companies might still be growing like revenue and profit at like 30% rate at like 5 billion revenue, you know, it's going to go away. So whenever the future is uncertain, that's when investors sort of, you know, pack up and leave. And I think system of record was kind of the thing that was supposed to stay. So everybody was starting to talk about systems of record. And are you going to be the system of record for whatever X, the thing that your customers are doing? And I think sort of Notion is sort of well-positioned to be almost like a generic system of record. Like you're going to have like your CMS system, CRM system is a system of record. And then you're going to have a few others. But for all the other like work stuff, like all the documents and things that you do, kind of need somewhere else. And if that's not, whatever, Google Drive, then Notion is like it, right? And Yeah. Just having, and it was sort of built as a better UI and could integrate between documents much better than a traditional, like Google Docs for example. So I think it kind of, it fits a really good generic horizontal use case. That. Whether it's so generic that it's going to be kind of eaten by like Anthropic and OpenAI's ability to create better UIs. Like, are they going to just create a, like, is Claude just going to be the new UI for everyone? Yeah. Maybe, right? But we'll see.

Rob Litterst (18:09)

Okay, I have a bunch of comments on some of that stuff, and I this. So I think like first and foremost, the challenge for Notion and like where I'm at in my head is like you've got these kind of like model companies that are building their own agent interfaces. And I think COG, like in my head, I think of like Claude, OpenAI, and Grok or GrokBot, which I've been pretty vocal about using as kind of like a layer above SaaS, right? Where like they plug in all my SaaS applications and they just become the kind of like interface where I can access those. And I think in that use case, Notion is an amazing product. Like I've been using Grokbot with Notion and like all of my agents push like shared documents and files into Notion. And we have like kind of archives of all this stuff that we've been working on. It's incredible. It's an amazing system of record for agents. And I think the challenge for Notion is like getting people over that hurdle in their heads that Notion is also an app where you can work with all of your other SaaS apps because it was a legacy SaaS application. I just don't think a lot of people think about it in the same way. I'd like the same layer as the models. So that I think is going to be a challenge. One thing about Notion is like, I think it was back in 2025, or maybe it was just earlier this year, but one of their founders was quoted in the Wall Street Journal talking about how their margins had already slipped because of AI. So people are clearly using their product for some of these AI use cases. Back then, I don't even know if like custom agents had been launched yet. So my assumption is that people were using the kind of like Notion Q&A agent to get a lot of stuff done and like automate a lot of the work that they were doing in Notion, which is great. This pricing change to me suggests that people have not been using their custom agents, right? Because previously their custom agents were their own kind of add-on that you needed to engage with by taking action, right? You had to go buy the custom agent add-on and start fulfilling it with credits. And my assumption now is that they were not getting a lot of people to bite on that or to upgrade on that. So now they're including these select credits, which you can use for custom agents in every seat as a way to try to drive usage. Those custom agents. But those are some of my immediate thoughts. Like I think Notion is getting a lot of AI usage, but their kind of fundamental challenge is whether they're a system of record, a system of action, or both. And I get the sense that like a lot of people are not actually using Notion as a system of action. I get the sense that people are primarily using it as a system of record. And any AI usage that they're doing is typically with the Notion agent. Just kind of doing work within Notion and not necessarily piping into like other products and tools using it more as kind of like an agent harness.

Ulrik Lehrskov-Schmidt (21:08)

Hey everybody, just a quick reminder to follow and subscribe to all of our channels if you like the podcast so you get all future episodes directly in your feed the moment they come out.

Rob Litterst (21:17)

Thank you and enjoy the rest of the episode.

Ulrik Lehrskov-Schmidt (21:23)

As you know, I'm writing a book right now on AI pricing. So The AI Pricing Roadmap, and I have this chapter on packaging AI, sort of figuring out, like mapping the value of an AI product. And fundamentally, I have this little two-by-two, which is like assets and actions, mine and yours. I'm calling it the assets-actions matrix because, I don't know, I'm good with words. I love it. I'm trying to sort of figure out something more, like, snappy. Like your assets is like proprietary value. So it's like traditional like software systems, like data infrastructure, whatever. So if I'm buying access to your storage, then, you know, that's it. I'm paying you to store my gigabytes in the cloud. And also just like having access to like the LLMs is something. And then you have there's like custody. So this is, you know, any system of record basically sits over there. Customer data, customer IP. But it's also things like for legacy systems, if you have a big SAP installation and you've just spent like two decades of like custom configuration of it, it's essentially like the configuration is yours. You pay for it, you pay maintenance fee on it, all these things. But you still have to sort of like an ongoing way pay SAP for the maintenance fee, even though it's your configurations, right? Right. So there's a lot of like monetization logic that essentially sits in like people paying for getting access to their own assets as they're being held by the vendor. I think what we saw was a lot of like broad strokes over the last decade. We've seen a lot of like monetization going from being sort of asset-related pricing, like my assets, to more being action-related. So usage or consumption-based was, hey, let's pay for the actual API call, the outcome or the whatever, the price per lead or the price per customer ticket that were resolved or whatever. And so I think the net new one is a little bit more trying to find good and new and innovative ways to monetize enablement and also to monetize like custody of customer data, even though the access is not being done by a human, but it's being done by an agent. So normally we could just price proceed and then the human would have access through their login to go in and like look at the data in the CRM system or in the marketing automation system, whatever it was. There's no longer a human now, it's an agent million times per second. And if there's only one agent and not a hundred humans, then there are some logics around how custody works that has to be changed. But it's fundamentally still custody. So whenever I'm trying to sort of map out, especially a product that is, I don't mean drifting as in like without navigation and destination, but more like that it is moving, it is undergoing change. So with like, are we still pricing the same thing or do we have to price something like net new in this matrix?

Rob Litterst (24:53)

So with Notion, you would say, so they've primarily been in the top left, I think, assets and yours, right? Like that's kind of like where they started.

Ulrik Lehrskov-Schmidt (25:05)

And I think now what they're seeing is like this bottom right thing where they have to price more on enablement.

Rob Litterst (25:10)

And that's where they're trying to drift, yeah. Along with these pricing changes, Notion announced a few other things. They revamped their model router, which they call Auto. This is kind of what they mean when they say like, when Notion picks the model. And this is primarily what you're going to be using those using standard credits for. So if you remember like down here, their are standard credits. That's going to help you with anything that comes through Auto. And then they also announced, Bring your own AI. So they have bring your own key. I think it's primarily from ChatGPT now and they're working on the others. And they announced a few other things like their improved MCP, AI command center. I think they just teased. I don't actually think anything is coming from that, but I would imagine that's kind of like what Ramp recently announced. They have their own kind of like AI HQ and control center called Glass. And so I think Notion is probably trying to kind of like follow But I bring this up because what you mentioned there, Notion has their own router. So as far as I know, I don't think Notion actually has like post-trained their own model. Like I don't think Notion actually has their own like AI model within Notion. But they have their own router that they can use to basically route things based on some combination of like quality, speed. I think that those are kind of like the two primary things that they're looking at. Or you can bring your own API. But as far as I know, there's no tax on that right now. Like that is just something you can do. And they obviously have their MCP as well. So right now, it feels like they are kind of like orienting around, like allowing you to do work in one place for simplicity, allowing you to do work cheaply by tapping into their model, and allowing you to do work kind of however you want to. So being like extremely flexible, right? So it's like convenience, cost, and flexibility. They're trying to go for everything, and it feels like they're throwing spaghetti at the wall. But based on what you said about enablement and that entire dynamic, I think that's really interesting with Notion. I recently read that Salesforce is actually charging people to bring their models into Salesforce or to use their models with Agentforce on top of it or something like that. So they are actually, I think, kind of pursuing an enablement fee in their pricing model, but it doesn't look like Notion is doing that. I would love to get your thoughts on it.

Ulrik Lehrskov-Schmidt (27:42)

Yeah, so I think it's you when you're doing pricing and also when you're looking at someone else doing pricing should always look at it as sort of as like a point in time, which is like, okay, so something came before this and something is likely going to come after this. I don't think necessarily that we should look at notions pricing right now and say, oh, this is their best take on how pricing for Notion should be. I think you can also look at it as, oh, this is a stage or a phase in a series of experimentation that will lead them to better pricing. I think you can look at, and even if you look at the time series that you showed before, like, hey, we're going to introduce AI, then we're going to start to redefine a little bit what it is, and then maybe gate it somewhere. Then we're going to use some credits. Credits, and then we're going to start to charge for them. It's essentially the same playbook that Figma did, which is also why I think I've referred to them in the beginning. But I think it's one of these where if you had to sort of put it in sort of a sequence, it's like launch, learn, monetize. So when you're launching, you're - well, you're not monetizing yet. So you essentially want to say, hey, can we get adoption of these products? And we are also exploring the value of the technology. So we're like making AI accessible in the product and we're not completely sure what the value is yet. So that's sort of the launch phase. You can do that if you have like a phase zero before that where you build some guardrails, like fair usage mechanisms, the right contracts, the right customer narrative. There's some things you could do. So it's like prepare for a good launch phase. But then you have a launch phase and then you do that. So it would normally be like expensive for models that you either subsidize, like give away for free, or you try to price at cost. And then a lot of this, like, undefined AI work starts to become repetitive. So, inside of Notion, there are probably like a handful of things that a lot of people do over and over and over again, finding missing data, completing, like, things. So once Notion then learns what these things are, it's like, oh, okay, a lot of people are doing these seven things over and over and over again. Let's now do two things. One, that's when you move it into monetize. We're going to define what they are and start to charge for them separately as outcomes and actions. It's not priced per like the underlying LLM cost. Like we can actually do credit-based pricing. We can do something where we just like, hey, five credits for this thing, 10 credits for that thing. That's how it's going to work. So we actually get to do a little bit of an outcome of value-based pricing on those particular items that we have now learned what are. And second, because they're so well-defined and repetitive inside our product, we can now fulfill them with a cheaper model. So we don't need the expensive frontier model. We can use like a cheaper open-weight model that can do it for a fraction of cost. So over this launch, learn, monetize phase, you have this margin expansion where revenue goes up because you get to price the actual outcome or action that customers perform. And then costs go down because you substitute expensive models with cheap models. When I'm looking at your Notion right now, I'm saying, hey, you're somewhere like halfway through like the first cycle of this.

Rob Litterst (30:58)

Yeah.

Ulrik Lehrskov-Schmidt (30:58)

Where you kind of introduce some AI, throw it at the wall, you're seeing what sticks, like you're starting to define some actions and some credit pools. And now you've also started to sort of differentiate between like expensive and cheaper models. And now you're sort of trying to move into higher margin consumption on the user base here, right? But they haven't sort of, I don't think they've sort of finalized it yet.

Rob Litterst (31:23)

Well, it's funny, like, I keep having this thought. Like, I've read a bunch about Amazon and, like, the business of AWS. And everything that Amazon did to generate better margins than the other hyperscalers. They did all this stuff under the hood. They built their own chips. They built all of these mechanisms that help ensure that they are driving as much consumption out of compute as they possibly can and make it as cheap for them as possible. And it feels like these dynamics are coming to SaaS, right? So like a company like Notion, like because they are serving consumers, I think at the end of the day, a lot of the products that they're competing with are going to end up being around the same ticket price. So it becomes a battle of like margins, which comes from like internal efficiency and kind of like routing and all of those sorts of things that I think the hyperscalers learned but now SaaS companies need to start to figure some of that stuff out. Like, do we post-train a model? Do we get really good at routing? Do we do both? Do we do like, what else can we do to make the costs way cheaper for us? So your point, I feel like Notion is probably going through that right now too.

Ulrik Lehrskov-Schmidt (32:33)

What do you think is going to happen next? Where do you think the friction is in this one?

Rob Litterst (32:37)

It's a great question. Getting back to kind of like the core takeaway from my article that I wrote this past Friday, just to give some context here. So I posted about this change before it happened and then got a bunch of feedback from that LinkedIn post. And so I was like, all right, I'm just going to write the newsletter about this week because I think this is kind of a good, like, evolving story. Story. So they introduced this two credit, this two kind of credit model balance, actually three credit model balance, Notion credits that kind of come in after the fact and like help you replenish both. And my broader idea is just that like, you know, if you have this one, like Notion again is not like a developer tool. It's not a tool where, you know, their customers are billing engineers and like understand like highly specific like types of credits and stuff like that. I think generally They are, their main users are not engineers for the most part and are not people that are used to like parsing pricing. Like engineers, I think are used to like breaking pricing down into like it's unit costs and like, you know, like figuring out like how much they need to use. And they're much more accustomed to that than I think other SaaS operators. And so this just felt like a very confusing and kind of complicated model for a company like Notion. And I hope to your point that it is kind of like a pit stop along the way to a clearer model. My general takeaway here is like having multiple credit balances, it takes the complexity of your model and it puts it on the customer, right? So it's like instead of you figuring out an elegant way to charge for everything that's happening in Notion that people can do within Notion, you are essentially outsourcing that complexity to your customers and like letting them deal with it, right? It's like, instead of just having one balance and a bunch of different ways that you can pull credits from that balance, whether it's using their router, your own tokens, whatever it is, you're basically putting together this pricing model that like makes all of that visible to the customer when I think you don't necessarily have to do that. So like that was my broader takeaway. And when I posted, so Sam Lee from Stripe, previously of HubSpot, and Snowflake mentioned, it might be technical constraints, so something that they would deprecate over time, pooled credits is technically challenging to implement, which I totally understand. Jeff MacMillan, who used to run pricing at Autodesk, chimed in. He thinks that it could be due to internal cost forecasting, but he kind of agreed with me. He's like, that shouldn't be the customer's problem to solve. And for one, it's where Sam agreed to. Sam left two comments, and the first one was like, kind of had like the bar face emoji, like, ugh, two credit pools. So yeah, I think like my broader takeaway was just that like there's this idea that I've been like thinking about quite a bit because I watched this video from Pat Grady, this partner at Sequoia recently. He talks about the diffusion gap and how AI is doing so much, but there are still so many people that are barely using AI. And so his whole idea is like this gap between AI capabilities and AI adoption is the entire opportunity. There's just this massive opportunity in the market right now to help people understand AI. And I also just read this book about NVIDIA and Jensen Huang. And one of the big takeaways is that Jensen Huang is like a teacher. He has like the Feynman method. You know how they talk about method? Like you don't really know what you're unless you can teach it to somebody really clearly and break it down and make it really easy. So Jensen Huang is incredible at this it's like is one of the superpowers. Like everywhere in the NVIDIA offices, they have whiteboards so that he can break stuff down. Other people can break stuff down. He basically tries to get the entire organization to be teachers that can make things really easy to understand. And so I'm going down a rabbit hole here, but like my bigger idea is just that like the companies that can kind of be like that can use the Feynman method and kind of like explain how they can get value out of AI and make it really easy to understand and also make the commercialization really easy to understand, I think are going to have a huge advantage. And Notion feels to me like a company that could totally be one of those companies that like shepherd is basically like the shepherd for the Sherpa for AI for their customers, right? Like helping them kind of like work through like everything as it evolves in the world of AI. But this feels like a kind of complicated first stab at this multiple credit model. So yeah, I would be curious to get your take on any of those rambles.

Ulrik Lehrskov-Schmidt (36:59)

I think you're right that from a strategic perspective, that's the opportunity they have, right? They have the opportunity that they have a large user basis and only some, but not all. And basically, the majority of them are not going to be like AI native because, well, nobody is. And I think they have been sort of the, let's say, the other system of record. Like your serum system, you have it over here, but everything else is kind of in Notion for the vast majority of Notion customers. It just makes it like a really fertile ground for AI to use. And if they could just welcome that with open arms and not too much complexity, friction, monetization, all these things, they could not only stay relevant, but become even more relevant for that customer base. And I think right now they're kind of tripping over their own feet. I think you're right that it is sort of internal like cost restrictions that are causing them to do multiple credit systems, especially as the credit systems aren't designed around different So, if you have Clay's credit system, for example, like you have one for data and you have one for action. So it's like, okay, that's clear. Two different things, data credits, action credits, and then I can kind of like double rail that. But here it's just like, it's everything, but is it like my actions or is it like your actions? So it's like, so, and that sounds like very much like a cost, like an infrastructure kind divide that has nothing to do with the use case as such. So I think what's going to happen is that it is going to be pulled apart by the market. They're going to have like, they're going to have so much friction. They're going to, and then one kind is positive and one kind is negative. I think they're going to have very positive friction. And then some customers just want to consume a ton more of these credits than the model is they're going to have to, like, whatever, chase the opportunity and figure out how to make room for that. And then the other one is going to be that people don't use all the credits inside of their seat models and that the customer base is going to demand fungibility, like, to, like, pull credits across all seats. So I think they're probably going to make some adjustments. They should make some adjustments to this within months, I would say. And I think the main reason they're not is that they don't have the right cost controls in place to be able to really subsidize adoption and run good learning in their customer base for this.

Rob Litterst (39:32)

I would not be surprised to see, like, the more I think about it, the more I kind of think that, like, I wouldn't be surprised if they end up with, like, a similar model to like the model companies like OpenAI and Anthropic where you have these like subscriptions and then you have like API pricing basically for like your enterprise customers like I would think that that's probably where they because the API customers they can get a way better margin they're like it's purely usage-based and your point like they're going to have these kind of like power law dynamics with their customers where some are just getting like so much usage and then they're going to have these smaller customers that just want simple pricing So it wouldn't surprise, like I feel like a company like Notion, the more variable the use cases your customers are using your product for, the more I feel like it makes sense for you to charge for tokens or like closer to, like index closer to tokens than outcomes. And so like I wouldn't be surprised if they end up going like closer to that side of the spectrum.

Ulrik Lehrskov-Schmidt (40:28)

For sure. I think having a very small like toll gate or like enablement fee and then just making like enabling different sort of levels of sophistication of like AI-native users to sort of use Notion as sort of the whatever hodgepodge system of record. I think the problem with a lot of these sort of database databases like Supabase and all these other things that a lot of employees and organizations are not really good at looking at databases. Like they need a nicer, cleaner interface. I think Notion can sort of maybe bridge both of those. I think there is a world in the multiverse where an AI-native startup goes out and says, hey, we want to build the generic system of record for all kind of knowledge work across organizations. And then we want to build an AI-enabled element layer on top of that. Got some traction for it and then got valued at more than Notion is today, which is kind of weird, right? Because Notion kind of already built those whatever 8 million end users that they have or probably more, So I think it's one where Notion kind of needs the right story in the sense that they need to sort of know where they're going and how they could probably be like way more relevant and have way larger ambitions that they ever had from their just, hey, we're going to be like better than Google Docs and cheaper, which was kind of their reason for existing like five years ago right and I think the whole action layer, they don't need to be just that. They can be, oh, we're going to be like a full-on orchestration layer for everything that has to do with AI for people that are not tech geeks. Yeah. And that business is way more valuable than Notion was or ever was in the past. That is it for this episode of Pricing Page unPacked.

Rob Litterst (42:23)

If it was useful, subscribe to the podcast on your podcast provider of choice, or find us on LinkedIn, the PricingSaaS community, or on YouTube.

Ulrik Lehrskov-Schmidt (42:30)

And remember, your pricing page is not just a page, it is a strategy statement. See you on the next Pricing Page unPacked.