In this episode, host Michael Bernzweig talks with Mitchell Jones, founder and CEO of Lava, about how to monetize AI features without destroying SaaS margins. Mitchell breaks down why AI introduces real variable costs, why seat‑based pricing no longer reflects value when agents can replace dozens of users, and how usage‑based models can better align price with outcomes. You’ll learn how an AI gateway plus billing stack can track per‑request cost in real time, enforce target margins automatically, and let agents both collect and spend funds so AI‑native products can scale profitably.
In this episode, host Michael Bernzweig talks with Mitchell Jones, founder and CEO of Lava, about how to monetize AI features without destroying SaaS margins. Mitchell breaks down why AI introduces real variable costs, why seat‑based pricing no longer reflects value when agents can replace dozens of users, and how usage‑based models can better align price with outcomes. You’ll learn how an AI gateway plus billing stack can track per‑request cost in real time, enforce target margins automatically, and let agents both collect and spend funds so AI‑native products can scale profitably.
Timestamps:
Michael Bernzweig (00:05.934)
I hope everyone enjoyed that last presentation. Coming up next, have Mitchell Jones. He is the founder and CEO of Lava. His organization is building next level generation payments infrastructure to power seamless AI native transactions and agent driven billing. So with that, Mitchell, I'm going to hand the mic over to you.
Mitchell Jones (00:31.673)
Thank you so much. How's everyone doing today? As as was said, my name is Mitchell. I am the founder of Lava. The short about me is I love payments in one form or another. My previous startup was a company that helped people save for retirement by getting their full 401k match if they couldn't afford to get it on their own. We gave you money to get the match from your employer. So if you had a five thousand dollar match, we would give you five thousand dollars to get it and we would get paid back through a
your 401k. So we were the only company in the world that figured out how to make the 401k a payment method for good. Before that I was at Facebook where I ran Facebook's digital wallet for Latin America and Asia Pacific and before that I was at Dropbox. But today we're going to talk about a big problem for almost everyone right now who is thinking about AI or using it in their products and services. The biggest issue right now facing traditional SaaS is the revenue model is breaking.
for AI. What do we mean by that? Well, first and foremost, AI introduces real variable cost. Before, in the previous era of SaaS, there were variable costs, but oftentimes, they were so low you could treat the variable cost as fixed. It was great for making simple pricing that was easy for customers to understand. You have a certain amount of seats. You pay a certain amount for them each month. The issue is now with AI compute, most
Pricing can get heavy usage can actually flip your margins overnight. A second issue is that seats no longer reflect the real value of your product. What do I mean by that? Well, imagine you're building an agent, right? And that agent can do the job of 20 seats. If you make that agent better and better and better, one day it'll be able to do the job of 30 seats, 40 seats. And what that means is you're actually making less money.
because you're building a better product. So seat-based pricing no longer reflects the real value and therefore your revenue streams are at risk. Finally, agents are both collecting money, but they're also paying for things. There's no real product that enables you to do both. And again, many of you may be asking, well, is this relevant? Is it important right now? Well, it is important right now.
Mitchell (02:56.42)
And the reason why is if you do want to monetize AI, it requires a new approach. You need to be able to measure your usage costs and your revenue. All of those pieces of information need to be right where you're paying and where you're getting paid. You also need to be able to have pricing that's value aligned. In today's era, one size pricing is not going to fit every single business model like it did in the seat-based fixed pricing era. And finally, you need to be able to have
some service that can both collect revenue but also pay for services as an agent is doing them. Most tools are unable to do that. You might ask yourself, well, why do I need to do this now? Well, most estimates say that there's about an $80 billion payments opportunity through 2030 for AI. About $4.3 trillion of spend will be AI spend by 2030. So this isn't something that most people can ignore and businesses cannot ignore it.
This has been a trend for years now. Back in 2023, about three out of five SaaS companies was trying to understand how to incorporate usage-based pricing into their product. You might wonder why were they doing that? Well, number one, one of the main issues that people find with seat-based pricing is it's not values aligned. If you're doing more work or more value, you should get paid for that. It's important for your margins, but it's important for actually
an important relationship and a proper relationship with your customers. And so what we do is we make it easy to incorporate the seat-based, subscription-based pricing of the past and the usage-based pricing of the future. We have two products that work together in order to help you do this. Any single tool that you're using from a legacy era of SaaS tools usually struggles when it comes to collecting payment and doing billing when you need to incorporate it in usage. We make it easy.
to have one API that allows you to make AI requests. So with the Lava API, you can actually access every single model out there, whether it's OpenAI, Claude, Mistral, 11 Labs, Base 10, you name it. From there, you get real-time analytics on how much it's costing you to serve customers. If an action or a product is costing you $20 per customer, we show you that in real time. What that allows you to do
Mitchell (05:21.644)
is then make your cost center your profit center. Because we show you that in real time, we also have payments orchestration that then helps you charge that customer the appropriate amount. If you want to make a 50 % margin, we make it very easy. We help you both understand your cost, but then also charge the customer. And we are the only company that not only helps you charge the customer, but also as you're collecting cash from those customers,
Your agents, if they're doing things that require them to also pay for other things, you can use that same source of funds for that work. AI is going to work both ways. Agents are real living. They don't just collect money in. They're also sending money out. And agents need to be able to earn revenue and pay for services autonomously. We're the only tool that starts all the way from routing API calls, on routing LLM calls, all the way
to accepting payment. And that one simple stack makes it very easy for any business that's trying to transition from a SaaS only model to a hybrid model, maybe where they have a subscription and usage-based pricing. And that's what we're here to help folks with.
What you unlock when your gateway meets your billing, and when we say gateway, we mean our AI gateway, which allows you to use us to route all of your calls, you actually now have an ability to have margin-aware pricing without a bunch of code. Most tools require you to have thousands of lines of rigid custom billing logic if you want to incorporate usage and AI spend into your pricing. We can do it in a few lines of code.
That cost attribution is tied directly to revenue. There's real-time credit enforcement and margin aware pricing. Whether it's our tool or any other tool you use, you have to be able to measure what you want to price. If you can't measure what you, you can't price what you can't measure. And really, you can't measure what you don't route, which is why it's important to have an AI gateway like ours that helps you route all of your LLM calls and payments orchestration tools that have you monetize on.
Mitchell (07:33.943)
Thank you. If you ever want to get in touch, we're happy to help you.
Michael Bernzweig (07:42.766)
Mitchell, I know that a lot of the founders and executives listening in today are probably suffering with AI, trying to figure out how to tie it all together. cost, obviously, with all of the API calls can add up very quickly if they're not monitored correctly. for a lot of the organizations that are dealing with this challenge,
When you first connect with organizations, what are the typical types of challenges that they believe they're suffering with, and what is the reality of what's going on?
Mitchell (08:26.283)
Yeah, most orgs really are in a state where they know they don't really have a tool that helps them upgrade their pricing. Right. Most organizations are stuck with a fixed subscription billing model and they are now incurring more and more and more AI cost. And so then they ask themselves, well, how do I upgrade my pricing in a way that allows me to both take in the subscription but also have some overage or usage attached to it?
They oftentimes try to look to legacy tools and those tools usually just don't get the job done. What we do to help them out is we make it very easy for them to really just route all of that traffic so that they can immediately understand how much it's costing them, but then also charge the customer for that. And we have the only end-to-end tool that helps you understand how much it costs and then immediately also charge them for that. Most companies are afraid or
are bracing for the impact of having to add payments tooling in order to solve this problem because they think it's going to be thousands of lines of code, thousands of lines of edits, and we can do it pretty cleanly in only a few.
Michael Bernzweig (09:38.234)
Now, is the challenge that each organization is working through the same or are there different types of challenges that you typically see for founders that are trying to solve this problem internally or on their own?
Mitchell (09:54.114)
Yeah. So I'd say it really depends on the company size. larger companies, the bigger issues usually around revenue recognition. How do I make sure I'm properly when I now have this usage based pricing or outcomes based pricing? How do I make sure I'm actually recognizing that in the proper form? Right. It doesn't come once a month. It's now coming when something happens. And most payment stacks were not set up for that.
And so now you have a tension between your CFO suite and your engineering team. The CFO suite often is like, hey, the current tooling we have doesn't really help us understand that. And it's important for us to be able to do our jobs and do this. The engineering suite has usually built something custom and they're like, I don't want to keep maintaining this, right? This is so much work to do. And there's a big tension there where we can come in and make it easy for engineers to integrate.
but also easy for CFO suites to get their insights. For earlier stage startups, it's actually a different problem. For them, it's I need to get up and running fast and I need to be able to accept money as fast as I can. How do I both do that and then also make sure if I'm building an agent, that agent can collect money, but then also let's just say it needs to pay for things on the internet as it's going in order to complete its service. How does it have a source of fun to be able to do that? What makes our tool so powerful,
is the same source of funds that the customer has funded that agent can then use in order to complete the service it's doing for the customer of that merchant. So it really depends on who it is. For larger companies, it's a lot of it's around how do I make sure revenue recognition plus billing works well. For smaller companies, it's how do I truly have an agentic payment stack? And how do I get up and running as fast as I can? I don't have a team of 10 engineers, right? It's just me and a couple.
How do I make that work and how do I not have to have a whole engineer dedicated to being our full-time payments engineer?
Michael Bernzweig (11:52.997)
Now for an enterprise organization, what does a typical stack look like in terms of what solutions are they trying to integrate the payments for?
Mitchell (12:05.174)
Yeah, so the typical stack for most enterprise companies, they have like a payment processor under the hood and then maybe some payments orchestration tool that helps them, you know, manage plans, do things like charging back and refunding the customers. And the big hole is once the executive suite decides, you know what, for us to continue making money as our seat model doesn't work anymore, we need to actually have a model where we're paying for the outcomes or the work that's been done.
Right. Well, you can switch that and you can also reflect that in your payment processor. But unfortunately, even if you do something custom in order to do that, there's still this big tension and issue of, sure, your product can do it, but you can't actually recognize it when it comes time to do reporting, when it comes time to do all the jobs you need to do in order to make sure you're running a compliant organization. And so, you know, the big thing that we tackle and we focus on
is really solving that tension. Engineers love us because it makes it really easy for them to do their job. C-suite teams love us because it makes it really easy for them to recognize revenue and actually do their reporting well. And most team stacks don't really have that right now. They have a processor, they maybe have a payment orchestration tool, and they maybe have some third party revenue analytics tool. But those are all disparate systems that the engineering team is forced to have to make work together. And...
When you add those three things plus custom usage based pricing code, this gets gnarly really, really fast.
Michael Bernzweig (13:40.33)
Yeah, exactly. So from the successful outcomes that you've seen, what does the journey usually look like from first recognizing the problem all the way through to getting your team engaged and getting the solution live?
Mitchell (13:56.065)
Yeah. So usually what happens is people come to me and they're like, hey, Mitchell, I've tried all of these legacy era payments processing tools, all these legacy payments orchestration tools, and none of them really do what I need it to do. And when they come to me at that point, I'm like, great. Well, let's actually get started. All you need to do is just tell me where in your code base are you calling AI. All you do is you switch out that one line where you're calling AI to those models and you add in.
our gateway in order to call those models. All of that cost tracking and data now flows exactly through your system. And then you just tell us how much money you want to make on each of those things and you set your pricing and your plans and all of that cost data will flow into your pricing. So let's say you said you wanted to make a 100 % margin on your product and your product has a $10 cost, right?
Or it's variable sometimes it's 10 sometimes it's 15, but you always want to make sure you have a hundred percent margin Well, we can do that very easily in our tool. All you do is you change those LLM calls and you switch them out with our router our API from there All you do is you tell us how what your pricing plan is how much you want to make on it and we do the rest
Michael Bernzweig (15:15.588)
Sounds like magic. we're looking, I'm sure a lot of organizations have some deeper questions. So if you were...
Mitchell (15:17.826)
There go.
Michael Bernzweig (15:25.002)
On the live event, you have a Q &A tab in front of you. Just type in any questions that you have for Mitchell. Coming up a little bit later in the event, he'll be on one of the Q &A panels and he'll be able to, we'll get through as many of the questions as we can. So, Mitchell, thank you so much for that great presentation and I'm looking forward to the Q &A.
Mitchell (15:47.971)
Perfect, thank you so much.