On this episode of The Innovators, I sat down with Mitchell Jones, founder of Lava, to talk about one of the least glamorous and most urgent problems in AI right now, getting paid without going broke.

Mitchell is building what amounts to billing infrastructure for AI products. If you are running agents, LLM powered tools, or anything where compute costs change by the minute, you already know the problem. Traditional payment systems were built for flat subscriptions. AI is not a flat subscription business.

Lava sits in the middle. It routes AI model calls through a gateway, tracks real time usage and cost, and lets companies price that usage in a way that actually preserves margins. You can choose how much you want to make per action, per credit, or per customer, and Lava handles the rest, including paying through to the underlying model providers.

What makes this urgent is something most founders are only now learning the hard way. In AI, your best customer can be your worst customer. Power users can quietly rack up massive compute bills while paying the same monthly fee as everyone else. That model worked in SaaS. It breaks fast in AI.

Mitchell’s insight is simple and hard to argue with. AI behaves more like a utility than a software license. Utilities are metered. SaaS pricing is not. Lava exists to close that gap.

We also talked about where the company came from. Mitchell has spent his career deep in payments, running Facebook’s digital wallet in emerging markets and founding a prior fintech company before starting Lava earlier this year. He did not wake up one day and decide to build an AI company. He talked to customers. Over and over. When everyone said they were duct taping Stripe together and hated it, he knew there was a real problem.

The conversation also veered into founder advice, especially for people outside the usual tech pipelines. Mitchell grew up in Dayton, Ohio. His path ran through finance internships, late CS coursework, Dropbox, Facebook, and then startups. His advice was consistent throughout, do not stare at the top of the mountain. Focus on the next step. Compounding effort matters more than pedigree.

Lava has moved fast. The company landed its first customers within months, raised a $5.8 million round, and now works with AI startups and legacy companies trying to shift from flat SaaS pricing to usage based models.

If you are building anything with AI under the hood and have felt that creeping sense of dread when the compute bill hits, this episode will feel uncomfortably familiar.

You can check out what Mitchell and his team are building at lava.so.