Transformation in billing: Metronome offers usage-based tech for the AI era – News

The story of Silicon Valley has been written by many engineers who got their start at a large tech firm, saw an opportunity to build something better, and started their own company to transform the way that business gets done. That’s the path followed by Metronome Inc. co-founder and Chief Executive Scott Woody.

Woody previously worked at the file hosting service provider Dropbox Inc., where he sensed an opportunity to shift away from subscription and seat-based models in the software-as-a-service world to hybrid usage-based approaches. Metronome’s pitch to investors and early users was that its technology could significantly reduce the engineering time and expense for billing systems.

“Actually making changes to the commercial model at Dropbox was technically quite hard to do,” Woody told News. “So the first principle of Metronome, in contrast, is to build a billing system that is hyperflexible, that allows you to price and package however you want in a self-serve way without writing any code. Change pricing in minutes, not months. That was really critical.”

As it turns out, Metronome’s approach is a timely one for the dawning new era of artificial intelligence. It’s still uncertain how enterprises can get a return on investment from their large outlays on generative AI technologies. Metronome’s service provides better visibility to enterprises on how much they’re spending on AI so they can better measure that return.

It’s picks and shovels such as these that could provide opportunities for the many other startups looking to provide generative AI services without the risk of competing directly with the AI giants such as OpenAI, Anthropic PBC, Google LLC and Microsoft Corp. or software-as-a-service companies such as Salesforce Inc. that have quickly embraced AI in their own services.

Shift to real-time billing

Woody set out to address another problem encountered in his Dropbox experience as well. When the price for a seat license changed in the middle of a month, customers didn’t realize the new cost until an invoice dropped on them at the end of a billing cycle.

“It would be like your power company changing your price on the third day of the month, but you only found out at the 31st, and you’re like, ‘Wait, what the heck?’” Woody said. “The fundamental technical issue is that the billing system wasn’t an online system. It was a batch, an infrequently updated thing. The way Metronome solves this is we do all of our billing and operations in real time, and then we provide the data from the invoice and the billing system.”

Users that pay to use OpenAI’s services probably don’t realize it, but when they check to see how much they’re spending on ChatGPT, it’s Metronome’s engine that’s providing the data. Metronome’s customers include AI provider Anthropic, along with enterprise tech heavyweights Nvidia Corp., Databricks Inc. and Snowflake Inc.

“If you’re an OpenAI customer and you sign into the dashboard, all of that data is Metronome data that’s just being fetched from our servers in real time,” Woody explained. “If you run a big job, you’re going to see that appear on your bill within seconds or minutes of actually completing the task. You always have a real-time view into your bill.”

Woody has described Metronome’s work as “reinventing billing with design.” Asked about the role that design plays in his usage-based model, the company’s CEO described how Metronome’s approach has contributed to a better overall experience for its customers’ users.

“A lot of our customers are building these product experiences that allow [their] customers to control their spend,” Woody said. “To me, that’s a design problem. It’s how you build a monetization experience for your product or service, that incentivizes customers to let you try and use all your products and services with the ability to sleep safe at night, knowing that they’re not going to accidentally spend a bunch of money.”

Need to control AI spending

This approach has struck a chord among users of AI-based services, according to Woody. The integration of AI into a multitude of product offerings has fueled interest in Metronome’s usage-based platform.

“AI is hitting every company on earth like a ton of bricks,” Woody said. “AI is really expensive to run. If you have a product or service that’s really expensive to run and the more you use it, the more you spend, you need to adopt a business model that allows you to capture incremental value.”

A prime example of how the changing landscape for AI is driving interest in different pricing models can be found in the growing use of agents, intelligent pieces of software designed to perform specific tasks. Salesforce has become a major player in the nascent agentic AI world with the deployment of its Agentforce platform this fall.

Agentforce’s pricing follows a conversation-based model, currently at $2 per AI interaction, with discounts for volume usage. Woody believes this will lead to an even greater need for monitoring costs.

“In an agent-based world, the value of Agentforce or Salesforce with agents is going to shift,” Woody noted. “It’s not going to be how many people have access to the data anymore. It’s going to be how many of these agents are running wild across your business.”

Tools for super high-scale data management

Metronome is a cloud-native business, built on the Amazon Web Services Inc. platform. As an online financial system, the company must process billions of events per day and not drop a single one. It relies on a number of open-source tools, such as the distributed event streaming platform Apache Kafka, to power its engine.

“Apache Kafka is our best friend, and it is one of these tools that allows us to horizontally scale up to accommodate almost any volume of incoming events with strong guarantees that we’ll never drop an event,” Woody said. “We need to be fully auditable as well. We generate invoices for usage records and invoices for hundreds of millions of end-customers a month. We need to be able to reconstruct every single one of those events or every single one of those invoices so that we could pass an audit.”

Along with robust record-keeping, Metronome also depends on technologies that can handle its streaming needs. That includes working with the data streaming platform Confluent Inc. and using Apache Flink, the open-source big data processing framework.

“A lot of the challenges we have are super-high-data scale problems, but more importantly, they are real-time streaming problems,” Woody explained. “In fact, our system is entirely architected around the concept of processing data in real-time, which I think is notably distinct from other billing vendors and also enables us to do everything that we do with super low latency.”

Metronome, named for the musician’s tool for keeping perfect time, was launched in 2019 and has attracted the interest of prominent Silicon Valley investors. The company’s Series A funding round of $30 million was led by Andreessen Horowitz in 2022 and it closed a $43 million second round of funding earlier this year, led by NEA.

Metronome 2.0 was unveiled in October. The latest features included an ability to launch new products with complex enterprise contracts and in-product billing dashboards to empower users with configurable spend controls. It’s part of Woody’s vision for Metronome as it seeks to reshape the billing landscape.

“We want to be the universal billing system that helps usher in this wave of consumption and AI based pricing,” Woody said. “That transition for these large companies is a multiyear business transformation. But I think with AI, it’s kind of inevitable that almost all of them are going to have to make it, and we want Metronome to be the leading light in that space.”

Image: News/Microsoft Designer