AI INTERNAL FRAUD INVESTIGATOR · GROCERY, HYPERMARKET, PHARMACY, CONVENIENCE

The case, not
the dataset.

AI internal fraud investigator for grocery, hypermarket, pharmacy, and convenience chains. Staff, and the suppliers who work with them. One machine at head office. Hours to a few days. Finished cases.

BOOK A SCOPING CALL OPEN A SAMPLE CASE
The POVA guardian raccoon
All the records
No store login
Offline at head office
Every finding tied to source
01 / ANATOMY

What a case holds

01

Named subjects

People, suppliers, stores, registers.

02

Money

How much. Which products. Which stores.

03

Chronology

The story, in order, with the timestamps behind it.

04

Relationships

Who is connected to whom, including the ties nobody declared.

05

How it was done

Which records it moved through: register, waste, deliveries, clock-in. And who was brought along.

06

Exact source proof

Every claim points to the exact file, row, and log line.

07

Competing explanations

The innocent readings. Tested, then kept or ruled out.

08

Decisive next action

What to do Monday. Interview, recover, escalate, or close.

And the audit of itself

Coverage, method, and what was left out. Written down.

02 / EXHIBIT

One finding, opened

Example. Not a real customer.

CASE 17-G · FINDING 04 SUPPORTED · ACT NOW
SUBJECT
Shift leader 4418 · Store 12
Supplier: “Harbor Dairy”
MONEY
186,400
WINDOW
11 months
RECORDS
2.4M register lines
+ WASTE · DELIVERIES · CLOCK-IN
CHRONOLOGY
AUGHarbor Dairy starts delivering to Store 12. Goods short at receiving, booked as waste after close. SEPWaste policy tightened. Waste stays high, on the same four nights of the week. OCTSupplier credit never filed. Register and clock-in put 4418 on the dock those nights.
COMPETING EXPLANATIONS
—Short-dated dairy going bad: waste rises only on 4418’s nights.
—Scanner error at receiving: the same supplier and the same short repeat.
→Goods short at receiving, booked as waste. Shift leader and supplier together.
SOURCE PROOF
register_store12.csv · rows 1,884,061–1,884,930 waste_adjustments.csv · rows 4,117–4,268 deliveries.csv · rows 902–1,043 clock_in.csv · rows 21,455–21,610
DECISIVE NEXT ACTION

Pull Harbor Dairy invoices against credits for Store 12, interview 4418 with the chronology, and stop deliveries from that supplier at that store.

03 / CASE FILES

Four cases

Examples. Not real customers.

GROCERYWASTE

Missing deliveries booked as waste.

Goods short at receiving, booked as waste after close. Same shift leader. Same supplier.

HYPERMARKETWAREHOUSE

Warehouse write-off hiding a short.

Stock written off in the warehouse. Same dock crew. Same supplier.

PHARMACYDRUG CAGE

Drug cage, waste, and who was there.

Waste log, who opened the cage, and prescriptions. They do not match.

CONVENIENCEREFUNDS

Fake-receipt refunds.

Refunds on one register. A loyalty card that only ever refunds, and it belongs to an employee.

04 / FILES

What you copy

Head-office files. Your team copies them. POVA maps them to each other.

MUST

Register sales (POS). Product list and store list. Waste, price cuts, stock changes. Supplier deliveries. Staff clock-in. Returns. Loyalty cards.

PHARMACY, AND USEFUL

Prescriptions, controlled drugs, cage, waste. Keys, cage, safe. Payroll. Gift cards.

ALSO JOINS

Operations. Audit files. Cyber and access logs. IoT and sensors. Other head-office files.

05 / METHOD

Give it the head-office files

01
Your team copies the folders. Register sales, product and store list, waste, price cuts, stock changes, supplier deliveries, staff clock-in, returns, loyalty cards. Pharmacy: prescriptions, controlled drugs, cage and waste. No live store. No cameras. No credentials change hands.
02
It runs. Hours, or a few days. It keeps going until every path is tried and every possible case is done. A week if it needs a week.
03
Open the cases. Read them in the POVA viewer, or export them into the system you already use.
ALL OFFLINE · ON PREMISE · AIR GAPPED · NO NETWORK ACCESS
06 / SECURITY

Offline, at chain head office

The analysis never touches the internet. One machine at chain head office, not in stores: POVA BOX, a sealed tower we bring, or a Linux desktop on a computer you keep offline. Same product. Your data stays in the room it started in.

+Air-gapped analysis: no telemetry, no error reporting, no background sync.
+No third-party AI in the analysis path. Your data never reaches an outside model.
+Deliverables are self-contained offline files. POVA retains nothing after delivery.
+The case registry can be delivered encrypted, readable only by the people you name.
+Deterministic analysis: same input, same result, with an embedded audit trail.
ISO 27001 · 22301 · 42001 · FULL SECURITY DOCUMENTATION AVAILABLE ON REQUEST
07 / FOUNDER

The person behind the system

POVA is built by Roy Daya, CEO. Certified Corporate Investigator (CCi). Certified Safety Incident Investigator. Data scientist. Kellogg MBA. Doctoral and post-doctoral AI research.

LINKEDIN →
IN THE PRESS
FRAUDBEATThe data detective companies call when the numbers don’t add up→

Bring a copy of the files.

First call. We look at what you already have.

BOOK A SCOPING CALL
You don’t need to prepare a deck.