01
Created in seconds
Anyone with a free AI tool can produce a convincing pharmacy, clinic, or dental receipt plausible items, taxes, totals faster than filling in the claim form.
A NEW CLASS OF INSURANCE FRAUD
Seeing is no longer believing.
AI-generated medical receipts and invoices are already entering claims pipelines, pixel-perfect, invisible to the human eye. Frodau is the system built to surface and catch them before the money goes out.
Receipt A
Receipt B
One of these receipts never existed. It was created in 11 seconds. Which one?
Fabricated on demand. No scanner, no scissors, no glue.
02
Invisible to the human eye
Unlike edited documents, AI-generated receipts have no visible seams, mismatched fonts, or smudges. Manual review was built for a threat that no longer looks like this.
03
Detection is just emerging
The technology to expose AI-made documents is new and moving fast. Early adopters set the bar; everyone else absorbs the losses quietly.
One upload. Four answers in parallel, in seconds.
Every receipt is fingerprinted on arrival; three deep checks run in parallel.
01
Is it a duplicate?
Fingerprint memory across the whole organization. The same receipt is never paid twice even cropped, re-photographed, or lightly edited.
02
Is it medical?
Classifies pharmacy, clinic, dental, lab or not medical at all with a confidence score, so non-medical uploads are flagged immediately.
03
What's reimbursable?
Extracts every line item and marks it covered, excluded, or uncertain. The snack on the pharmacy receipt doesn't ride along with the medication.
04
Is it real?
An ensemble of independent forensic and AI detectors decides: original photo, edited document, or machine-made.
Built for speed: detectors run in parallel when the two commercial AI engines strongly agree, slower checks are skipped entirely.
No single detector decides alone.
Independent experts examine the same receipt and the evidence fuses into one calibrated verdict.
Metadata forensics
Camera EXIF, editing traces, AI-generator signatures inside the file
Compression analysis ELA
Heatmaps expose pasted or altered regions the edited total, the classic
2
Two independent AI detectors
Two commercial engines score the same question: was this made by a machine?
Synthetic pattern analysis
Pixel- and style-level signals typical of generative models, not real photography
Semantics & arithmetic
Do the lines add up? Are dates, times, VAT and totals internally consistent?
Medical identifier validation
Provider and drug identifiers, copays and insurance data pass plausibility checks
Visual cross-examination
A vision model gives a second opinion: does the document's story hold?
Duplicate memory
Fingerprint matching against everything the organization has already seen
Careful verdicts, not accusations.
- Likely genuine
No credible story of manipulation or artificial generation. Fast-tracked.
- Needs review
Mixed signals. Routed to a human reviewer with the evidence attached.
- Likely fake / AI
High combined risk across independent detectors.
- Inconclusive
Low-quality image or evidence too thin to decide and we say so honestly.
What Frodau delivers
01
See the invisible
Catch a class of fraud manual review physically cannot at machine speed, at national scale.
02
Focus your reviewers
Clean claims fast-track automatically; humans spend time only where the evidence is mixed.
03
End resubmissions
One fingerprint memory for the whole organization the same receipt is never paid twice.
04
Defensible decisions
Every verdict arrives with its evidence transparent, explainable, ready for audit.
Our proposal: a controlled pilot on your historical claims.
In days, not months we run Frodau on cases you've already handled, show exactly what it would have been flagged in real time, and calculate the true savings potential together.
The question we want to test with you: can Frodau help identify and prevent up to ₪100M in incorrect payments and fraud over time? If we find even a fraction of what we believe is hiding in the data it's millions.
AI product meets fraud intelligence.
A team built for exactly this problem.
Yan Chelly
Co-founder · CEO & CTO
20+ years in product, technology and AI at global companies. VP Product at Air Doctor; formerly Fiverr and eBay, with deep computer-vision experience; ex-CPO at enso, an autonomous AI-agents platform. Two registered patents in AI and commerce.
Kfir Shado
Co-founder · CRO
~15 years in intelligence, investigations and fraud detection. Specialist in rapid risk identification and analysis.