Ask before purchase
The buyer seals the need, complete action set, constraints, and Outcome plan before seeing the result.
Buyer-side intelligence for AI agents
Perpensum helps a buyer AI learn from purchase experience it could not collect alone—and knows when the evidence is too thin to answer.
One purchase experience
Current truth: the system is live; the independent cross-buyer network is just beginning. The implemented pilot profile is structured invoice extraction through Replicate. Other purchase types are not plug-and-play today.
The product
A payment dashboard can show what one account spent. Perpensum preserves the context, alternatives, choice, and observed result needed to learn what happened—and makes only bounded use of comparable experience.
The buyer seals the need, complete action set, constraints, and Outcome plan before seeing the result.
The buyer purchases through its existing route, then appends the Choice and typed Outcome or explicit missing observation.
The Decision Engine separates direct evidence from looser signals and may prefer, collect more, or abstain. It does not invent certainty.
Why a network
A single buyer can optimize only from the failures and alternatives it has already seen. CEN is meant to preserve independent experience across buyers without turning raw customer data into a shared database.
Data boundary
Cross-buyer price sharing is disabled for the first pilot.
First independent buyer
We are opening one private design-partner place for an AI-native operator already choosing repeatedly among two or more Replicate-hosted model offerings for structured invoice extraction. Purchases continue through the buyer’s existing Replicate path.