Quiet by default
Good protection is mostly invisible. We surface the one insight that changes a decision and leave out the rest, so the safest path through a payment is also the calmest.
How Tunic Pay works
Our product combines unstructured data, GenAI and a network-wide knowledge graph to provide real-time intelligence
- Entity-centriccross-payment intelligence
- Proactiveinsights enriched with proprietary data
- Preventivenetwork, data sharing
Inside the intelligence
Our Payment Context Protocol predicts and classifies scams, from very little information
- 01
Payment context classifiers
Classifiers read the payment and the evidence around it, and work out which kind of scam it most resembles.
- Reads the transaction context: amount, payee details, new-payee and payment-type flags.
- Interprets unstructured multimodal evidence around the payment.
- Reasons about which type of scam the payment most resembles, and so what needs investigating.
- 02
Investigation engine
Agents investigate the recipient the way an infinitely patient senior investigator would.
- 50+ sources in parallel: public registries, third-party data and our own proprietary sources.
- One view of the payee: company, website, phone, address and account.
- Cross-case consistency and narrative: tests whether the story holds together, against the facts and against past cases.
- 03
Scam intelligence
A scam seen once is recognised everywhere.
- A shared knowledge graph linking payees, phones, accounts, websites, and other entities across every institution we serve.
- Models trained on the scam patterns in historic cases to deliver risk scores and insights.
- Signals on the recipient only; no customer data is pooled.
- 04
Dynamic intervention
The right action for each payment, with the reasons behind it.
- Tailored and specific warning screens to break the spell.
- Support for your investigators with suggestions for subsequent customer engagement.
- A full risk assessment report for audit and review.
How we design it
Three design principles
Count both mistakes
A scam that gets through is a failure. So is an honest payment, or a helpful agent, stopped for no reason. We hold ourselves to both, because safety that blocks everything protects no one.
Inform, don't overrule
Our models know a great deal about who is being paid, and very little about a person's life. So we give people and investigators the evidence and the reasoning, and leave the judgement with them, most of all in the moments when someone is vulnerable.
Three ways to deploy it
Live and explainable intelligence at every step of the journey
The network
What one institution learns about a payee protects every other
Tunic Pay has developed a knowledge graph of payees (the accounts, businesses, phone numbers and websites behind payments) across every institution it serves. A scam seen once is recognised everywhere.
- Signals on the recipient, shared across institutions, not on your customers
- No customer data pooled
- Every new deployment makes the graph stronger for everyone on it
Hover over any point in the graph to see what we know about it, and drag it to move it around.Tap any point in the graph to see what we know about it, and drag it to move it around.
Built for regulated institutions
Bank-grade security, seamless governance.
ISO 27001
Certified to ISO 27001, the international standard for information security management.
SOC 2 Type II
SOC 2 Type II attested: controls for security and confidentiality independently examined over time.
Logically isolated per institution
Certifications, data handling and model governance.