Follow the dependency graph of a transaction.
Ask any spending query.

About the project

Managing spend policy at companies is hard. Filing taxes and passing IRS audits is hard. We fix that. apriori answers two questions before money is spent: Can I buy this? And why?

An employee describes a purchase in plain English, like "Client dinner Thursday with Acme's CFO, about $48," or pastes a product link. They get one of three answers:

Allowed. A card authorization is issued for exactly that amount. Not allowed. It names the exact rule that was broken and the smallest change that would fix it. One more detail. It asks the single question that decides the answer.

How it works

A SystemOne (Jev) turns the sentence into facts, each with a confidence score. Anything it is unsure about stays unknown instead of being guessed.
The Z3 theorem prover checks those facts against every rule at once. The answer is a proof, not a prediction, so the same purchase always gets the same answer and every "no" cites its reason.
Purchases that need sign-off go to a manager on the dashboard. Approved ones get a mock card authorization carrying a hash of the proof, then land in a QuickBooks-style ledger that the same rules audit in full.

It's faster than regular AI prompts, always consistent since it is backed by proof-based logic (needed for financial data), and incredibly cheap (in our 24h of testing, we used up 3 cents total of our Jev budget).