Kyle Poyar x Jared Robin | Revenue Creator

All right. This is another episode of the Revenue Creator podcast and newsletter. I'm joined with Kyle Poirier. Kyle, welcome.

Thanks for having me, Jared.

Do you want to give a quick overview on your background and what led you to creating this monetization report and some of your insights?

So I have been a pricing nerd for more than 15 years. I actually started my software career doing pricing strategy consulting for tech companies. This was in 2010 when we were still working on on-prem to SaaS transitions, if you can believe that. And, you know, honestly, SaaS pricing, while it's always been important, it's been a little bit boring. Like the mechanics are kind of we'll charge per seat or maybe we'll do platform fees, we'll start cheap, we'll raise prices over time as we launch more products. Like once private equity gets involved will really jack up prices, right? Like there were only so many things that you could do and it was, it was something that was pretty straightforward. I think what's fascinating with AI products is they can be extremely disruptive because the value of AI products are very different from traditional software. Many times AI is doing work on your behalf as opposed to being a platform for people to log in and do work. So products are more of a system of action. As opposed to a system of record. And that means we can't necessarily rely on the old ways that we used to charge for products. But at the same time, like, we got to sell to CFOs, right? We've got to get buying approval from procurement. And customers are still pretty comfortable buying seats. And they're not as comfortable buying with models like outcome-based pricing or usage-based pricing. For the most part, the reason why I created this report is like, finally, this is a really interesting time for modernization. If feels like we're all struggling to figure out what the path looks like for both SaaS and AI products, and I wanted to get some objective data about what's happening.

Yeah. What intrigues me the most about pricing is whenever there's an inflection point or like a major categorical shift, there tends to be a new pricing with that, right? Like from on-prem to SaaS, from buying your DVDs to getting them streaming, right? Like, and it tends to go hand in hand. With, I don't want to use the word moat, but like in creating something new, right? Like, because it adds an extra barrier to the old way. Like on-prem couldn't do a per-seat license. It maybe technically could, but like the way it was structured was maybe impossible, just like AI gives you the ability to do more usage stuff and et cetera, et cetera. So like, that's where I geek out on it is that it typically joins An actual disruption, right?

Yeah, I think that's, that's spot on. And, you know, candidly, we haven't settled on what this model looks like for AI yet. Based on the data, the number one path that companies are taking right now is hybrid pricing, which is a combination of subscription pricing, usually seat or platform fees, and then with consumption on top, which is usually in the form of AI credits. So hybrid pricing is now the most popular pricing model. About 2 in 5 companies have hybrid pricing models, and a lot more are planning to introduce some sort of AI credit or token model. And I'll share my data here. So the data specifically shows that 29% have some sort of AI credit or token model today, which is, I mean, Up over 100% from last year, right? So this is a pretty big increase, but still it's less than a third of software companies. 33% said that they don't have AI credits, but they plan to introduce them within the next 6 to 12 months. So expect to see a huge wave of companies doing this. One example is actually GitHub announced that they're going to launch an AI credit model starting, I believe, in June. Figma very recently rolled out credits, I believe just in March. March of this year, and so a lot more credit models. And the driver is actually not so much AI-native companies; the driver are large SaaS companies that are doing more than 50 million ARR. That's the group that's in particular that's sort of leading the charge around the credit model. And Satya Nadella from Microsoft actually said this really well in the latest Microsoft earnings call. The way he sees modernization with AI is that they're shifting to a seat model to seat plus. Consumption business model. And for their customer service product, actually, he said over 60% or roughly 60% of customers are already purchasing these usage-based AI credits. And I think that what was fascinating with the earnings call, if you listen to the whole thing, is he says that like seat-based pricing isn't going away. It's just kind of becoming arbitrary. So they're still selling seat licenses. It's just seats are essentially an entitlement to some consumption. And then you could buy extra consumption on top of that. But people still like feeling like they're buying a seat, feeling like there's predictability. And so Microsoft's still going to sell seats, but under the hood, the pricing is increasingly credit-based.

So seats is almost semantics for just having a minimum amount of credits per login.

Exactly. And so it gives people a feeling of predictability. Like you can kind of license the way you're used to, Licensing, and then, you know, you'll ultimately pay more in future years as folks consume more of the product.

Why do you think— I think I know the CEO of Microsoft said otherwise, but can you see a world where it's completely consumption-based?

So I go back and forth around where I see things going. I think that these AI creditor token models are really challenging for buyers. It's one thing if like You know, a couple vendors are doing it, but if you've got 15 different vendors that all have different credit models, and by the way, they define credits differently. They have a different margin that they're taking on credits. So some are really expensive and some are fairly cheap. Some are bundled into your seat-based subscription. Some, you know, you pay for purely on a usage basis. Oh, and plus you're already actually buying tokens through the AI models. Themselves, you probably have a relationship with OpenAI and/or Anthropic. So if you're like an enterprise team, like, how do I make sense of all of this? How do I route my spend? And I'm actually— so I think that VCs actually are really pushing for usage-based models right now. They want their portfolio companies to be in the token path. Token usage is growing exponentially. So any business that monetizes based on tokens just has this like natural tailwind for the business. So VCs are pushing for it, but I'm not sure buyers really want that future. And I think they're going to really struggle with it. And I think a lot of them are going to say, hey, wait a minute, why don't we bring our own tokens? So we'll kind of plug in our API keys and that way we'll just, we'll have one master agreement through Anthropic or OpenAI, and then we'll sort of use those tokens across the application layer. And if there is the shift towards that model. And if buyers really push for it, then the vendors will actually have to move away from kind of the classic usage-based pricing and find more value-based pricing models.

Years, years ago, we're talking 2012, we actually did an outcome-based— it worked in e-commerce where we just took a percentage of revenue uptick. It was interesting. And I don't know if we could go in that direction for B2B from a revenue standpoint, because so much is dependent outside the tokenized system, right? Like with the humans involved and stuff. But the learnings and how the organization was constructed could lean towards today in the sense that the CS function was handled by product people versus CS people, right? Because they were trying to maximize the JavaScript, different elements tagged to understand who's going to buy and not buy right now so that more people bought. I think there are some pretty cool takeaways. And then also there probably are outcomes that can control, that, that can be controlled more like meetings booked or something to that effect. And when you do that, the unneeded token consumption, like the company is going to put more levers to not allow people to do anything unneeded and spend recklessly. Whereas now we don't have that.

Right.

You could, you could build whatever workflow as inefficiently as you— heck, Claude Code might guide you in that direction to use that. So like, I know there's a lot of levers there, but if everybody singularly focused to drive value and then also there needs to be a standardization of one company to the next of some sorts. Like I signed up for company, 150,000 credits, $200 a month, and that got me 6,000 emails and I'm like, What? And then another company, 150,000 credits might be 150,000 emails. The general consensus is like, it's like half that because it needs to do a thing to get the emails, you know?

And it's just like, well, it's so impossible to compare across vendors.

Is that your inefficiencies as a company or is that my inefficiencies? And frankly, It might be in the middle or leaning towards the company, right? In those cases?

Well, that's, yeah, to your point, with any sort of usage or consumption-based pricing, the vendor's often saying like, I'm going to pass on risk from myself to you as the customer, like, use as much as you want, I'm protected, right? There's, there's not gonna be any unprofitable customers, because you're going to be paying based on consumption. If you end up letting your intern use Opus, or maybe Mythos when it comes out, and they keep them meter running and they rack up a $10,000 Anthropic bill over the course of a week, like, that's your fault, that's not our fault. And so I think with this hybrid pricing, there will be incentives for vendors to try to add more sort of spending controls, visibility. I mean, AWS has really pushed in this direction. If you remember 10 or so years ago, there was a cottage industry of optimization players like CloudHealth that would give sort of visibility and control over your AWS bill. And promise to help you shave it by 10 or 20%. I think that the AWS eventually realized that we have to add this in app, like we have to give users this level of control in order to really scale the way we want to. And so I do think AI companies will have to go in that direction. But for now, the incentives are really not aligned with them doing that. Without company pricing, it takes that off the table because the underlying infrastructure costs don't, don't matter. The customer isn't paying for some sort of arbitrary consumption metric, they're paying for some sort of business value that they're getting. And so you as a vendor take on the risk, but you also can have a lot more pricing power because you're instead of pricing based on something that feels like a commodity and a race to the bottom on price, you're pricing to something that has a known business value to the customer. That they're, you know, ideally willing to pay, pay for, like they're willing, you're willing to pay a lot more for a meeting booked with a qualified prospect than you're willing to pay for a token, right? Just semantically sounds way more valuable. And so when I ask companies what was their expected pricing model in 3 years, there's a bifurcation. So about half said they expect hybrid pricing, and then the next biggest bucket was outcome-based pricing. And so we really see this like this is the kind of core tension within AI pricing right now. So what I love about AI pricing, it's like such a disruptive model, or such a disruptive message to your customer. It really shows that you're willing to stand behind your product and the business value it's going to create. I think it's a much better sales pitch, and it really connects to the business case because customers finally believe in the business case of your product versus it being sort of that like sales collateral that gets thrown, you know, at the end of a deal.

Now, what types of outcomes would it be based on? You know, outcome-based pricing, because, you know, revenue means the most to people, but it's also the hardest to connect the dots to, right? So like, what's, what's reasonable? And one quick thing I noticed from outcome-based pricing was, I'll tell you, our ICP got very, very fine-tuned to who we could actually drive outcomes for. Right? So where we were a young company and we were like, oh, let's do SMB, they quickly got off. And then when we landed somebody that was doing $10 million a month in revenue, like a Home Shopping Network thing, and we showed them $1 million net new and we took 8% of that, that $80,000 bill they did run from. That was scary. But it was objectively shown, and that's $80,000 a month. How many people are making $1 million off a $100 million client? Even very, very few. They're probably making $250,000 or less. So there is the potential for way more money, but there, oh my gosh, what a process figuring that out though. Psychologically, it's tough to be honest.

Yeah. Big potential upside, but a lot of risk on both sides. And As soon as you start charging for outcomes, the customer starts looking at the bill really closely. Like, did you drive that outcome? Or did we drive it with maybe a small assist from your product? And to your point, the outcomes—

third-party analytical tool, we had to use a third-party analytical tool and have a real-time holdout group to show.

Yeah. I've definitely seen that. And I think to that point, like there have been performance-based or outcome-based models in different spaces, right? Advertising, You know, used to not be outcome-based, like used to pay to get a placement in a newspaper or a magazine or on a billboard. Now you're expecting to pay for performance, whether it's on Meta, Google, or elsewhere. And then you can in some ways define the performance metrics that you want to pay for based on your objective for the campaign. I think that's maybe an interesting direction for outcome-based pricing can go. Also in the recruiting space, right, you're going to pay a lot more for a contingent search, where you're paying based on, you know, success fee as a percentage of maybe first year salary, you're gonna pay a lot more for that model than you're gonna pay to post a job on Indeed or LinkedIn. But it's because the, the vendor is delivering a service and, you know, their neck's on the line, right, if they don't deliver that outcome. So in tech for AI products, the outcome could be a million things. I think that's part of the challenge is that You really want to only offer this if there's consistency in what types of outcomes your customers are looking for. It's come, I think, the fastest to the customer support space. And in that space, they have kind of a, like, somewhat binary, although still fuzzy definition of when AI can resolve a customer support interaction without it being escalated to a human. That's something that, you know, that's a success. That, that's something we're going to charge you for, we're going to charge you $0.99 per resolution. And so I think that was a really disruptive message. Intercom was early in launching that with their Fin product, they now rebranded their whole company to Fin. But I think even in Intercom's case, so they initially were $0.99 per resolution, they now actually say that their pricing is $0.99 per outcome. And they've started to broaden what an outcome means. Because if you have a— let's say you have a customer that comes in that has a sales request, they want to buy something more from your product, but they go through customer support. If you resolve that through AI, you're actually missing out on a lead, you'd probably rather have a successful handoff to the right person within the company with the information gathered about the account. And so I think what Fin is saying is that there's actually multiple different outcomes that are valuable to you as a business that you should pay us And it speaks to kind of that outcome-based pricing tends to get more complex over time as people see these edge cases, as there's variety of outcomes, variety across the customer base and so on.

But yeah, ironically, at the same time, there's now a holistic solution for all aspects of go-to-market. And that's super interesting. What they're doing though, with the outcome on the handoff, that's something interesting to Discuss, because like that is reasonable. It should be reasonable, something along those lines for any go-to-market team, right? Or I'm sorry, when I say go-to-market team, I mean go-to-market technology.

Exactly. Like, I think that those are justifiable things to charge for. And they have— they also now have plenty of tools for businesses to help improve their outcome rates or tailor what kind of outcomes they're looking for. And I think their philosophy is that we're extremely customer aligned and The more we can help you automate, the more you're going to save as a business, and you know the more money we're going to make. And so now all of a sudden, sort of product usage and success becomes that North Star around revenue growth, as opposed to some arbitrary token consumption. And like it really doesn't matter for the customer whether a support interaction involves like a thousand tokens or one token; they just care about like resolving that customer interaction in a way that has. You know, fast response time is available 24/7, ideally in the language the customer speaks or writes in, right? Like has a has a great customer experience associated with it. So I think it's much more of an aligned model. Now, what's one thing I do find fascinating about the space is that one of Fin's biggest competitors is an up-and-coming startup called Decagon, and Decagon actually offers customers two pricing models. So they say we'll do outcome-based pricing like Fin, or we'll give you a more predictable pricing model that's a price per conversation that gives you budget predictability over the course of the year, as opposed to some sort of variable outcome-based number. And I think what's interesting is like, it sounds less customer-friendly, right, to do that model because the vendors, like, they don't have skin in the game on the outcome. But what Decagon has said publicly is about, I believe, about 80% of their customers choose the more predictable conversation approach to pricing because they value predictable pricing that much as it is so much easier for them to budget for. And customer service is often a cost center as opposed to a revenue driver in a lot of companies. So this kind of speaks to me too. Outcome-based pricing is a fantastic message to the market. And I think for every company, if you charge on an outcome basis or not, you should think about what things would look like if you did. Because that's going to force you to really get a lot more clarity in what your product does for customers, how to talk about the business case and how to set up a North Star for your product and customer success teams at an account level. But you might not end up charging exactly on an outcome basis.

The market needs to be educated from Fin's point of view on why this is actually better for them, do you think? Or because it sounds like companies are giving what CFOs want based on what they were trained to want and predictability. Whereas if they could articulate this new way to being predictable or predictable enough, it might be more advantageous for them.

Yeah, there's absolutely a need for better education here. I mean, I think about this as in the on-prem to SaaS model, in many ways, like the SaaS model was so much better for customers. Like you pay way less upfront instead of software being a capital expense, it's now more of an operating expense. You can fire the vendor if they're not performing. There's now an expectation of things like customer success that didn't really exist in the on-prem world. Software companies actually started innovating more and like introducing new features and you wouldn't have to, you know, buy a whole new product just to get access to one feature. It was just in your plan. Yet this transition from on-prem to SaaS was very slow. It took years and years and many customers still preferred to buy in the old way. And so this education effort will take time. And in some ways, like companies have to maybe even rethink how they budget for technology, right? Software companies are normally pulling from technology spend, but AI companies maybe need to pull from vendor spend, or maybe it's even headcount spend, right? Other kinds of pockets of capital. And so this is, we're still early here. And I think that part of the next frontier for AI companies is teaching customers how to buy your product, because every vendor is going to now say that they do the same thing. Or you could even build something yourself with Claude code, etc. And so vendors are going to have to show why their product is better, even though on paper, the products can all do the same thing. And so I think this is actually going to need a lot of buyer education about how to buy AI products and how to evaluate AI products against each other. With pricing being one of those factors, along with a bunch of other things.

This has been awesome. Super insightful, Kyle. There's nobody that's dived as deep as you in this. I'm excited for the report. How could people see the report and get in touch with you?

So I have all the key insights on Growth Unhinged. You could just go to growthunhinged.com or Google Growth Unhinged, and it's right on the homepage.

Thank you for your time, my friend. This has been awesome.

Thanks for having me, Jared.