The back-office employee who never opens a report — because it already read them all. Five live anomaly detectors screen every delivery and flag the costly mistakes with the money at risk attached, pushed to your email and messaging apps without anyone visiting a dashboard.
In most kitchens the numbers are checked by hand, long after the stock is through the door. Deliveries, invoices, usage and takings each pass through a person and a spreadsheet — and every handoff is a place an overcharge or a wrong-unit receipt can quietly slip past.

Stock lands and the paperwork piles up in an inbox.

A clerk eyeballs a few lines against a spreadsheet — most go through unread.
spot-checked only
A price creep or a wrong-unit receipt clears approval and gets paid.
overpaid
The gap surfaces at month-end — if anyone reconciles it at all.
found too late
By the time it is spotted the cash has left — no way to claw it back.
A wrong-unit order — cases booked as pieces — puts months of stock on the shelf in one drop. Spotted by hand, it is signed for and paid before anyone does the maths.
A quiet few percent on a staple, month after month, never triggers a second look. Nobody is comparing this invoice to the last one — so the creep just compounds.
Costs drift up while the menu price stands still, and an item quietly turns loss-making. Without live costing against 12 months of sales, it sells at a loss for weeks before it shows.
OperaNova replaces the manual spot-check with an always-on cost controller. Five live checks screen every delivery and sweep usage, sales and margins — and the moment something is off, it pushes an alert with the money at risk attached.

Every stock delivery, invoice, usage and sales line read the moment it arrives — nothing waved through.

Wrong-unit, price spike, usage variance, sales anomaly and loss-making margin — each scored with the money at risk.

A clean email and a message to your chat apps fire in seconds, duplicates filtered — no one has to open a report.

Every flag carries severity, the money attached and a full audit trail; cost and margin trends surfaced on the dashboard.
Human-in-command: the five detectors and the recipe advisor flag with the money attached, and the agent acts through an approval gate — a person always makes the call.
Catches a wrong-unit order — a delivery far larger than it should be — before it is signed for.
Flags a supplier charging above your recent average, so the quiet creep never compounds.
Reads stock used against the recipes, and takings against their 12-month norm, to surface the gap.
Watches live costing against menu price and warns the moment a dish turns loss-making.
No AI Autonomous Digital Employee is "done" until it hits an agreed target on your real work. Here's the shape of the contract we'd sign for this role.
Illustrative example. Your real baseline, target and acceptance criteria are set together, from your data, before any build begins.