Getting AI monetization right has become a prerequisite for both AI-native and established software companies. Both are grappling with the best way to meter usage, manage costs, and set prices that make sense for customers while remaining sustainable for the businesses. They’re also working out how to evolve their monetization infrastructure to support how their pricing strategy is changing.
The conventional wisdom around whether to build or buy your business software is fairly simple: if you’re trying to solve a problem that is core to your product and may prove to be a differentiator, build it yourself. If you’re trying to solve a problem that has nothing to do with your core product, buy a tool and move on.
But in the age of AI, monetization infrastructure doesn’t sit comfortably on either side of this binary. Pricing has become a fundamental part of product design: it shapes how products are built, how they go to market, and how customers use and derive value from them. Companies now ship pricing updates as quickly as they ship new features—and the pace of both is accelerating.
It’s hard to argue that building and maintaining a billing solution from scratch is “core” to an AI company’s offering, or the most effective use of scarce engineering resources. At the same time, finding existing tooling flexible enough to accommodate this pace can be a challenge.
Increasingly, companies facing the “build-versus-buy” question are choosing a third, hybrid option. They’re elevating pricing to a strategic product function, and allocating the engineering resources required to manage it. But they’re not building the stack from scratch: they’re choosing flexible, composable infrastructure that lets them keep their team focused on the decisions that matter.
Here are two common scenarios we’ve seen as companies evaluate how to approach their pricing infrastructure—and how this hybrid approach manifests in real life.
Mismatched pricing pushes startups to move quickly
AI companies can misprice in two fundamental ways: by pricing below the cost to deliver their service or by not fully monetizing the value customers receive. We’ve worked with companies in both positions:
|
Chipp.ai, which helps businesses build and run custom AI agents, originally priced its service on a traditional SaaS subscription model: $29 per month. But customers routinely ran up usage costs that far exceeded what they were paying—one user generated $4K in token costs in just a month. Chipp’s pricing became a risk. |
Clay, which provides sales intelligence software, also initially priced on a traditional subscription model. However, customers were deriving outsized value: before joining Clay, Everett Berry, now the company’s head of GTM engineering, said he routinely closed $80K deals using a $300-per-month subscription. |
These pricing mismatches can be particularly problematic for early-stage startups, where runway is precious and a rogue power user can erode it quickly. At the same time, engineering resources are especially constrained, and pricing fixes may come at the expense of product ships.
Buying a foundational billing infrastructure helped Chipp move with the speed necessary for their product and company stage—but that doesn’t mean they relegated pricing to a back-office function. They still designed their custom pricing structure, closely considered what margins to charge on tokens, and preserved flexibility for more iterations in the future.
As AI coding tools get more sophisticated, building a billing stack end to end may seem more viable even for very early-stage companies. If one engineer can vibe-code a solution, you’re also not pulling as many people away from the core product. But maintenance and future adaptability remain open questions for these kinds of systems. It may be easier to build a billing system that works for a company’s current circumstances, but building something flexible and low maintenance long term is still a challenge.
When companies grow more complex over time, billing infrastructure and engineering resources become strained
Many companies transition from simple subscriptions to some form of usage-based billing to resolve the kinds of mispricing described above. A 2025 survey of approximately 2,000 global businesses found that 92% of AI companies with usage-based billing models had changed their pricing model at least once since launching it. Elena Verna, the head of growth at Lovable, recalls making 10 pricing updates in her first year at the company alone.
This means that getting your pricing system right once isn’t enough. Companies need infrastructure that’s capable of evolving as they grow and launch new products. Retell AI, which sells AI support agents, initially launched with an in-house system for tracking the variables they needed to send accurate invoices. But as they grew, that approach quickly became unworkable: their team of 16 spent far too much time handling invoices, while invoice errors eroded customer trust and the lack of auto-billing resulted in high rates of missed payments.
Acquiring the right foundational infrastructure, as Retell eventually did, can make it possible for a small billing team to handle the needs of a fast-growing company. ElevenLabs, which has undertaken rapid expansion of both its geographical footprint and its product line, has managed to get this done on Stripe with just a single billing engineer.
ElevenLabs didn’t outsource monetization or treat it as secondary to its product strategy. It continuously iterated its product and launched new monetization models in concert, including a new usage-based credit model that charges based on the exact number of AI tokens consumed.
At the same time, it built on top of infrastructure that absorbed some of the complexity of rapid scaling: an integrated tax engine to support new geographies, integrated analytics to analyze performance, and a revenue recognition engine to handle accrual accounting. This foundation prevented small pricing changes from becoming full-scale engineering projects. Engineering time went toward supporting the pricing decisions unique to their product, not toward reinventing invoicing and payment collection.
How Stripe and Metronome can help
Foundational billing infrastructure isn’t meant to turn pricing into a back-office function. It’s meant to help companies maximize the impact of their engineering resources by ensuring they’re allocated to the pricing and product features that actually make a difference.
Stripe provides the most complete monetization infrastructure in the market, so you can charge and manage customers however you want while collecting and recovering more revenue, automating workflows, and accepting payments globally.
Whether you’re getting started, scaling, or transforming your monetization model, every need is covered: Stripe Billing powers subscriptions and recurring revenue, while Metronome, a Stripe product, powers usage-first and hybrid pricing models and sales-led contracts—all on a unified platform with native support for payments, invoicing, tax, and revenue recognition.
If you’d like to discuss your pricing strategy, and the infrastructure needed to support it, reach out to our team.
Additional resources
Learn how Stripe and Metronome are building together:
Read more monetization stories from industry-leading companies on Stripe: