How it works
From a paper invoice
to an approved action.
One loop, eight stations, one human authority. This is the entire product — there is no other magic.
Invoice
AI parses · people confirmPaper, photo, or PDF — invoices enter the moment they arrive. AI parses vendors, lines, quantities, and prices.
Review
Human checkpointStructured review puts a person between parsing and consequence. Each parsed line is confirmed or corrected once.
Ingredient Master
Reviewed truthVendor language resolves into canonical ingredients with alias memory, normalized units, and correction history.
Inventory and recipes
Derived · labeledConfirmed costs flow into quantity on hand, available servings, expiry windows, and every recipe that uses the ingredient.
Economics
Deterministic mathDeterministic rules recost plates, recalculate margin floors, and price waste risk in dollars. No model guesses a number.
Action Card
Bounded · sourcedWhen a signal crosses a threshold, it becomes one bounded, source-backed recommendation with reason and estimated impact.
Approval
Manager decidesA manager approves, edits, rejects, or ignores. Nothing consequential runs on its own. Rejections teach the system.
Audit
Measured outcomeEvery decision writes a permanent receipt; estimated impact and measured outcome stay separate columns forever.
Walkthrough
From scan to same-day action.
Seven steps, one operating loop. Scroll to see how each screen connects to the next decision.
Scan paper invoices or import PDFs and move straight into review.
Parse warnings
Lines (2)
Vendor, date, totals, unmatched rows, and confidence states surface before anything is saved.
Parse warnings
Lines (2)
Vendor language turns into canonical ingredient records, aliases, units, and costs.
Lots, quantities, expiry dates, and value at risk become visible.
Ingredient changes flow into recipe cost, available servings, floor price, and margin status.
Suggestions stay bounded. Managers approve, edit, or reject.
The dashboard prioritizes the next action for the day.
Good morning
Protect margins before service.
What needs attention now