"Tell the board we put AI in our billing system," Johnny exclaimed in the Slack chat. Rocket emojis abounded.
The board then patted themselves on the back and bragged to their VC buddies about how they now had AI in their billing system.
In fact, they went so far as to say they probably wouldn't even need an EMR anymore. The future was 100% AI. The EMR could just disappear.
But Johnny and the board had overlooked two things.
1. Generative AI is probabilistic.
And US healthcare billing contains some of the deepest rabbit holes imaginable, with edge cases where two completely different answers can both sound perfectly reasonable. At sufficient scale, your AI will handle the same situation one way today and another way tomorrow. That's not exactly ideal when money, claims, and ledgers are involved.
2. We're living in an era of subsidized compute.
Compute is cheap right now because somebody else is helping pay for it. Get everyone hooked first. Figure out the economics later. Now build a billing system processing billions of transactions, attach a token cost to every decision, and add the percentage of cases that need to be reconsidered, rerun, or corrected because the AI behaved differently. That adds up quickly.
So when your AI billing system reaches the point where it costs more than you can afford, come talk to us. We'll hook you up with our patent-pending, rules-based billing system where we use AI to help build the system, not to be the system.
And sure, you can tell yourself token prices and compute will keep getting cheaper. Maybe they will. But you're betting that those savings will outpace the exponential growth in AI usage by you, your company, and the rest of the world.
If you're a private practice and don't want your billing left to chance, come check out Ambiki.
And if you're the EMR leader or VC from the top of this post, send me a DM. We're considering an early-adopter API.