Anti-counterfeit with AI vision — the veteran's eye, on every box.

The problem

Human checking gets tired; the counterfeiter does not.

Professional counterfeiting is no longer street-stall knock-offs. Professional counterfeiting:

1. Replicates the packaging with unsettling precision — the seal, the neck, the tab, the generic hologram. It fools the untrained eye, and it fools the trained eye after eight hours of looking at boxes.
2. Reuses authentic serial codes — bought on the parallel market, reprinted, cloned in blocks. The serial OCR says "it exists in my ERP", and it still is not yours.
3. Comes in through reverse logistics — grey return, repacked, redistributed. The least watched vector and, in premium brands, the most expensive one: your own channel puts back on the shelf product you never made.

Classic checking (operator + magnifier + checklist) works 99% of the time. That 1% is the one that reaches the shelf or the end customer — and each unit is brand cost, not just hard cost. A margin eroded by parallel market does not show up in the P&L until it has already happened.

How it fits the IRIS system

iLEAN does not add a fourth system — it gives eyes and memory to the line you already have.

The problem with anti-counterfeiting is not missing data: it is that the human eye does not scale, and the serial information lives on an island separate from the product image. iLEAN acts as the filler that closes that gap, without asking you to change your packaging line, your distribution centre or your logistics provider.

Edge reads the product's visual signature. Connect reads the serial code and the expected destination. The agent cross-checks against the ERP and assembles the dossier. Legal and brand sign off the withdrawal — never the other way round.

The three iLEAN pieces applied to anti-counterfeiting:

• Edge — a terminal with machine vision (CNN) on the packaging line, at the reverse logistics intake and in the regional hub. It reads the product's visual signature — neck embossing, print screen, alignment, ink — and fires the actuator (ejector, stack light, hold) in milliseconds. It works with no network. If the centre loses WiFi, Edge keeps reading and holding boxes.
• Connect — captures the serial code, the delivery note and the return manifest, whether they come from a modern ERP, a vertical WMS, the importer's email or the distributor's messaging app. When a brand exports to several countries and each one reports differently, Connect normalises from second zero.
• Anti-counterfeit agent — lives in Central, inside the three safety rings. It cross-checks Edge's visual signature against the serial read by Connect, against the issuance recorded in the ERP, against the authorised channel and against the expected destination. If something does not add up, it assembles the dossier — before/after photos, serial number, batch history, chain of custody, applicable regulation — and hands it to the legal and brand owner. The person signs off — the withdrawal never executes on its own.

See the full IRIS architecture →

Before and after

Manual checking + isolated OCR vs. cross-checked vision with iLEAN

Aspect	Operator + serial OCR	With iLEAN Edge + Connect + Agent
Detection of professional counterfeits	Depends on a trained, well-rested eye	Visual signature against your own pattern, on every unit
Cloned serial	"It exists in my ERP" = it passes	Serial + visual signature + channel, cross-checked
Parallel market / grey return	Invisible until it reaches the shelf	Caught at the reverse logistics intake
Time from detection to legal dossier	Days, with manual searching	Minutes, with the file closed
Coverage per shift	Drops at the end of the day	Constant 24/7
Operation with no network	n/a	Edge keeps running on cabinet power

Impact estimate

Impact estimate for your plant — to be validated with your numbers.

The block below is an estimate to be validated with the specific data of your brand. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.

• Premium brand (cosmetics, spirits, OEM spare parts, consumer electronics) with documented counterfeit product on marketplaces and growing grey returns in reverse logistics.
• Edge pilot on the packaging line + reverse logistics intake. First expected value in a few weeks: the first counterfeit or parallel unit held with a closed dossier.
• Indicative payback between 4 and 9 months, depending on affected volume and % of share lost to the parallel channel. A reduction of counterfeit/parallel product getting past the check of ≥ 30% is expected, against your current baseline.
• The hard lever is not only the cost of the withdrawn product: it is the market share recovered in the authorised channel and the minimum price defended.

And the brand director's fair objection

"What if the AI flags an authentic unit as a counterfeit and we block a legitimate shipment?" — hallucination is a problem of free generation, not of anchored tasks. In tasks where the AI only compares an image against your own pattern and cross-checks a serial against an ERP, the best models pushed the error below 1.5% ^[1]. And even so, no withdrawal executes on its own: the dossier goes to the owner and the person signs it off. The three safety rings are there precisely for this.

^[1] OpenAI paper "Why Language Models Hallucinate", 2025 — on the reliability of AI in anchored tasks.