🦄 Making great presentations more accessible.
This project enhances multilingual accessibility and discoverability while preserving the original content. Detailed transcriptions and keyframes capture the nuances and technical insights that convey the full value of each session.
Note: A comprehensive list of re:Invent 2025 transcribed articles is available in this Spreadsheet!
Overview
📖 AWS re:Invent 2025 - Why Usage-Based Billing Must Be Treated as Infrastructure (AIM109)
In this video, Austin from Metronome explains why usage-based billing must be treated as infrastructure in the modern AI era. He discusses how monetization has evolved from licenses to SaaS to hybrid and agentic pricing models, requiring real-time processing rather than monthly batch billing. Metronome solves the "pick two" problem of correctness, low latency, and high availability by achieving all three simultaneously. The platform processes business events in real-time for invoice computation, entitlement gating, and fraud detection. Austin highlights how Metronome powers companies like OpenAI, Anthropic, and NVIDIA, noting that OpenAI made 10+ pricing changes in twelve months through their platform. He emphasizes that 85% of software companies are adopting usage-based pricing, making billing infrastructure critical for product-led growth and enterprise scalability.
; This article is entirely auto-generated while preserving the original presentation content as much as possible. Please note that there may be typos or inaccuracies.
Main Part
The Evolution of Software Monetization: Why Usage-Based Billing is Infrastructure
Hello everybody. I'm Austin. I'm here from Metronome, and we're here to tell you why usage-based billing must be treated as infrastructure. So this is actually our first AWS re:Invent that I've been to, and I couldn't be more proud to be here. We have folks from across the industry, companies like Salesforce that invented SaaS to modern startups like OpenAI and Anthropic defining the future of what computing is.
So I'm Austin. I'm a Solutions Engineer here at Metronome, so I've helped develop and land and expand some of our largest clients as well as our startup program, which I'll tell you more about towards the end of this presentation. Metronome does support some of the largest names in AI as well as usage-based software today. So those are names that you might have heard of like OpenAI and Confluent, as well as Anthropic, NVIDIA, LangChain. It's been pretty fun walking around the expo floor this week and seeing a lot of the logos that we've helped support, grow, and grown with over the past several years.
So why must usage-based billing modernization be infrastructure in this modern age? I think we're all familiar with the previous eras of software. You sell a license, you kind of forget, and they use as much or as little of that service as you provide. Then we moved to SaaS, subscription, consumption, seats, a little bit more of this value match. But as you can see, let's try out the laser clicker, that's fun. We are here at the nexus of this new era, this agentic outcome, workflows, actions, hybrid, you name it. We are seeing new pricing models be invented nearly, I wouldn't say every day, but it's pretty common, and they're all coming into Metronome and discussing with us about how to meet the value of their product with their end customers.
And it's also safe to say that Metronome is here and has been here to support these pricing models, things that we like to call hybrid, which are really exciting. So as companies transition from pure seats to seats with allocations, trying to understand and meet that value with their end customers while also understanding that the underlying cost structure of their service is changing. LLMs are obviously pure usage, and so when you actually use them, you're burning down on something, you're incurring costs. It's very difficult to match that towards a seat or subscription, so we're moving in this new age where, as I'm sure you're seeing from the entire expo floor, everybody's talking about agentic pricing. It's very difficult to just slap that onto a subscription and say that you're done.
Monetization is now real time. So in the past, billing used to be this kind of slow humdrum service. You'd batch processes, you give your customers a monthly invoice. They're all backward looking. You have to update things in spreadsheets. It's all kind of relegated to the back office. Finance doesn't really care or talk to anybody else. It's really about collecting bills at the end of the month, and those subscriptions are really what they are. They're really detached from the product experience and detached from that value that we're talking about.
Monetization now is real time, things like your in-moment workflows and notification systems, gating entitlement, as well as making it core to the product experience. Everybody here on this expo floor can certainly tell you that if you go and use ChatGPT, you're interacting with a thing called tokens. Now, can everybody describe what a token is? I'm not sure. Some days I don't know, but it is core towards the usage of these products. It is a strategic growth lever as you're connecting the experience with your customers to the end state of actually using the product.
So monetization is the infrastructure, right? It is the continuous flywheel of developing your product and continuing to iterate and match with your customers. So Metronome in particular does this through what we call our events. These are business events, any sort of transaction that might incur costs, that might incur usage, that might change the utilization of your product. These are things like an API call. I keep talking about LLMs. Inferences being made, servers spun up, spun down, continuous heartbeats while a service lives. These are all things that we're very familiar with and have been around for years, but actually attributing to them in real time and doing things like gating off access, letting customers know they're going into overages, or even just from a go-to-market perspective, letting them know, hey, you've hit 90% of your pre-commit, you might want to engage in an upsell.
These are things that we think are critical, and they can't just be done on even a daily basis or even worse, a monthly basis. These have to be real time. These are engagements that are pretty critical that we're all starting to become very deeply familiar with whenever you hit a little bit too much send on Cloud Code and it runs for a little bit too far.
That's the kind of real-time product experience that you need, and that's why this is an infrastructure problem. This isn't just something that happens in the back office anymore with invoices and revenue data. These have to be understood in real time, and this is also where Metronome starts to sit. The usage aggregation and the invoice computation happens on every single event. It is generating a new invoice. That's how we think about these sorts of problems.
Solving the Transactional Observability Challenge: Metronome's Infrastructure Principles
So why is this hard? Why is Metronome even around? Usage-based billing is not some novel concept. We didn't invent it. We just seem to be getting it right. That's because in the past, transactional observability was a pick two problem. You had correctness, low latency, and high availability, and everybody just kind of chose two and said it's good enough. But again, this is extremely difficult.
Correctness on its own for financial services means you need to make sure that every single cent is attributed and matched. You need to make sure that you're ready for your audits as you continue to be, let's say, a public company like Confluent, or you're a startup making sure that you actually pass an audit so you can raise that next round of funding. Events also may arrive out of order. These things need to be burned down on the actual attribution of that timestamp that you're sending into Metronome. This can't just be a situation where we'll receive it eventually and it gets attributed. Correctness is key towards a billing platform.
Low latency is again what we keep talking about with those real-time service alerts for entitlement gating and fraud detection. We're moving towards these usage platforms where customers are PLG oriented, so customers have immediate access. But as a business provider, you don't necessarily trust somebody who might just be signing up for your service. Now you might give them a free trial credit, but the difference between that and actually making sure that you're not running and incurring bad debt while also growing your business is essential. Your customers are also starting to expect real-time information. There are plenty of platforms that we're all very familiar with that maybe give you a 24-hour update after usage. That can't be the same thing for a PLG style consumer-oriented product. You need to understand in real time what your burndown is, and also if you're talking about a finance team, they need to understand across all sorts of customers, all sorts of users, where they're actually at month to month.
And then that high availability. This is because it's infrastructure, because it lives at the core of this product, you need to have extremely high availability. Billing can never go down. Your product might be able to go down, but billing has to be available. It has to be ready to be up at every single moment. Those dashboards that power all sorts of these things have to be extremely available so you can understand where your clients are at. Again, why is Metronome what we are? Because we got all three.
The infrastructure principles applied to monetization are these four. It has to be scalable. It has to be modular. You have to be able to orchestrate things, and it has to be extensible as your company and your customer base grows and as you iterate. Nobody actually gets pricing right on the first go, so that means that every single one of these pieces of this puzzle have to be extensible and have to be scalable. You have to be able to change things. You have to be able to think ahead as well as potentially backdate things and extend the platform as your company grows and you start to build more pieces of your billing stack.
Scalability, for one, is real-time information on ingest. That's invoice computation, as I talked about, during usage. Surges can happen, but you need to be able to ingest in real time and not fall over. OpenAI, as I'm sure you can imagine, scaling up and scaling down all in a very flexible fashion. Modularity is really where Metronome got things right over the past couple of years. Decoupling every single piece of your billing model so you can do things like having extremely nuanced customer contracts for your largest customers who are really good negotiators, and then the opposite side of things when you have a PLG segment that is completely separate but also the same products at their core. Being able to roll out those pricing changes while not being able to do things like migrations. We like to say that migrations are a dirty word at Metronome. We don't do them. Everything is updated in real time.
And that orchestration, that is things like having those invoices be sent off to your actual payment provider, having those dashboards inform your customers, those spend alerts. You can have things like a 90% threshold warning, let customers go into arrears but still gate them off when they hit 150%. These are becoming core product experiences that are essential and deeply entrenched in your billing, as well as those credit ledgers. You need to understand at the end of the day, for revenue recognition purposes, for burndown purposes, where things are actually going.
And that extensibility. Metronome sits at the heart of your entire monetization.
We've talked about those usage events and transactions that are coming in. They're feeding into this platform that has real-time metering and rating, pricing and packaging, which is extensible and flexible, as well as contract lifecycle management. This includes knowing when your customers are going to be ending their contract, when to engage with them again on a sales motion, when to move them to a free trial, and when to move them to an arrears contract.
From there, you get what's shown on the bottom and the right-hand side of this graph. This includes your CRM as well as your billing portal and all the data lakes so that your team can understand extremely deeply where your finances are at and where your customers are at. You can evaluate, understand, analyze, and iterate on your pricing and packaging, as well as manage the right-hand side where all the money actually goes. That's what we at the end of the day actually care about, because while we're doing a consumption-based business, we need to make sure that we're collecting the invoices for our customers. It isn't just for the love of the game. It's actually about making sure that our business is healthy and that those invoices get sent somewhere. And of course, AWS Marketplace, Metronome also sits and provides for them, as well as a little company called Stripe.
Preparing for the AI Era: OpenAI's Success Story and Getting Started with Metronome
You have to be prepared for this velocity of AI. AI is not just the buzzword anymore. It's the businesses that we all operate in. Companies are testing their pricing models and making iterations. OpenAI in particular made ten-plus pricing changes in the past twelve months. Somebody might wake up and decide that they want to compete on a different model and they drop the price. They do that through Metronome as well. Eighty-five percent of software companies are adopting usage-based pricing. It might not be core and essential towards their pricing model, but it might be some hybrid aspect as they are rolling out things like an agent on top of their platform or integrating LLMs into the product experience. These are becoming essential towards the new era of software.
Modernization engineers, this is a new title that we're seeing actually on hiring pages. They should demand all these aspects out of their platform. Can your service process millions of events in real time? Can you handle that product-led growth scale that is the new way to grow your business? It isn't just a marketing scheme. It is actually how you mature customers and early adopters into enterprise customers. Can you stay up with everything? Billing needs to be essential and needs to always be available. Can you deploy those pricing changes without real deploys, without migrations? Can you do it with a few clicks from somebody who is not an engineer?
Can you set this pipeline up to be completely removed and let RevOps then do things like pricing changes and pricing experiments in your platform? Can finance and product share one source of truth? This isn't just an engineering problem. This is a finance problem. This is a BizOps problem, but it is something that can be solved by engineers evaluating the proper tool that will take you from today through the end state of potentially this entire AI era. Again, Metronome powers this modernization infrastructure, so we sit at the center of this with real-time invoice compute and a modular pricing model so we can create, iterate, roll through pricing changes, and issue out those usage statements and invoicing data.
Obviously, OpenAI is pretty essential towards our business, and we have a pretty good case study for them. It's on our website if you want to read more, but Metronome is at the heart of how OpenAI operates. We actually like to say that when OpenAI joined us, we got the name wrong. They were some random little research lab we called OpenAPI. They're now a little bit more relevant towards our day-to-day lives. They're adding new LLMs every day. They are deploying to millions of users in real time. They're making pricing updates that are not months-long discussions and committees. These are done with finance, and they're matching the value while understanding the analytics that are coming out of Metronome, and they have ultra-efficient billing operations.
A little exciting thing that we're doing here at re:Invent: if you want to come by our booth and sign up to claim one million dollars in customer billing credits, you may have to go through another demo, you may not, but we will get you set up and live, and you can start using Metronome today, playing around, kicking the tires, and get again a million dollars worth of customer billing credits with Metronome. Our booth is right over there, somewhere near Cribl-ish. There's a number on there, but nobody's actually looking those up. Thank you. Again, come see us at our booth.
; This article is entirely auto-generated using Amazon Bedrock.
Top comments (0)