'Termination Was Immediate': Worker Processed 40% More Refunds Than Colleagues, Then AI Revealed Why

A pharmacy worker was paying out refunds for products that never came back and pocketing the cash, a scheme traditional checks missed until an artificial-intelligence tool flagged 40 per cent more refunds than colleagues. The worker was dismissed on the spot.

The case was recounted by Chris Kirksey, chief executive of the US digital-marketing firm Direction.com, which manages search optimisation for pharmacies and also oversees their inventory and loss-prevention systems.

Speaking to Forbes, Kirksey said the pharmacy fills around 400 prescriptions a day and handles dozens of refunds each week. Most are routine returns from customers who change their minds or receive the wrong item. The business had installed an image-capture system to track high-value stock such as brand-name drugs and costly supplements, after repeatedly finding gaps between the recorded count and the shelf.

Three months in, the software flagged one member of staff. Every refund that person filed was checked against the image logs, and the same gap appeared each time. The employee scanned the refund barcode, but the camera never recorded any product going into the return bin.

Kirksey reviewed 27 refunds over six weeks. In each one, no goods came back, yet the cash drawer still reconciled because the refund was booked as store credit rather than cash. The worker pocketed the equivalent sum, a running total of $2,400 (£1,790). Shown the footage, the employee admitted everything. 'Termination was immediate', Kirksey said.

Without the image logs, he added, the shortfall would probably have been blamed on supplier errors or customer theft. The same setup now runs across all three of the company's locations with no repeat incidents, and it is piloting a module to track stock levels in real time.

Kirksey did not name the pharmacy, its location, or the worker, and there is no indication the matter went beyond dismissal to the police or the courts.

AI-Faked Receipts Surge Across US Workplaces

The case surfaces as AI reshapes older, more familiar forms of workplace fraud, with expense claims the clearest example. Research by expense-management firm Emburse, released on 23 June 2026, found that 40 per cent of US employees surveyed had used AI to generate a fake receipt. Of those, 19 per cent fabricated a purchase outright, and 15 per cent inflated the value of a genuine one. Among UK workers, the figure was 29 per cent.

The survey of 2,000 staff was carried out by Atomik Research. It also found that 40 per cent of those who faked a receipt used AI tools paid for by their employer, and 9 per cent had coded their own. Emburse tied the behaviour to money worries, with more than a quarter of US workers, 27 per cent, admitting they had passed personal purchases off as business costs because of their finances.

Detection figures point the same way. Data from the audit platform AppZen showed AI-generated receipts made up 70.8 per cent of the fraudulent receipts it flagged by mid-May 2026, up from zero in March 2025. That tally covered 1,471 fake receipts submitted by 745 employees across 174 companies, claiming a combined $148,143 (£110,700).

Where AI Monitoring Meets the Law

The shift is pushing more employers towards AI monitoring, and with it a set of legal limits. Bogdan Condurache, co-founder of the website builder Brizy.io, told Forbes that such systems must stay tied to specific business risks rather than becoming blanket surveillance of staff. The point matters for smaller firms, where one tool can end up watching everything at once.

Katie Maguire, an employment partner at the law firm Devonshires, said monitoring in the UK has to meet tests of necessity, proportionality, and transparency, with workers informed under GDPR and the Data Protection Act.

An AI flag, she added, is a starting point rather than proof, and tribunals still expect employers to back suspicion with verified evidence such as system logs, receipts, or CCTV footage. 'A single AI alert is not enough to justify disciplinary action.'