Scam report triage for bank fraud teams
Victims report a scam in a short chat. Rastro builds the evidence packet, links it to earlier reports that share an account or phone, and ranks which network to act on first.
- Networks, by priority
Builds the evidence packet
A PDF for the victim's bank, ready in under a second.
Links shared accounts and phones
Reports with one account, phone or link become one network.
Ranks networks by urgency
Fraud teams see the network to act on first.
Logs every step it takes
Each step shows if rules, SERV or a person did it.
One platform from first report to ranked network
Someone called me pretending to be from my bank's fraud team and told me to move my money to a safe account.
How much money did you send, in US dollars?
I sent $2,400.
Collects the report in a chat
Victims answer in their own words. No form, no account.
Links reports that share details
An account, phone or link joins reports. A similar name never does.
40 reports paid into the same receiving account.
Each claim points to the links that prove it.
Explains why, not only a score
Every claim in a brief is checked against the graph before anyone reads it.
What it did on the demo data
These numbers come straight from the engine's audit log.
- reports read
- 0123456789
- networks found
- 0123456789
- seconds from report to packet
- .012345678901234567890123456789
- analyst hours saved (estimate)
- 0123456789
Hours saved is an estimate: 0.75 hours per packet and 2 hours per network brief, if built by hand.
Benefits for fraud teams
Less time stitching reports together. More time on the networks that matter.
Open the consoleActs on the worst network first
One shared account across 40 victims outranks ten unrelated reports. The ranking follows the money, not the order reports arrive.
Shows the evidence behind each claim
Every statement in a brief points to the links that prove it. Select a reason and the graph lights up.
Leaves the decision to a person
Rastro never freezes money or says anyone is guilty. Doubtful cases go to review instead of being guessed.
How it works
One report goes in. A packet and a ranked network come out, usually in under a minute.
- Step 1
Collects the report in a chat
The victim answers a few questions in plain language. Rastro asks only for what is missing: the amount, when they paid, how they were reached and the account they paid.
- Step 2
Builds the evidence packet
It writes a PDF for the victim's bank with the story, the amounts and every account, phone number and link in the report.
- Step 3
Links and ranks the networks
Reports that share an account, phone or link join one network. Fraud teams see the biggest and most urgent network first, with a brief that says why.
Two ways in
If you were scammed
Report it in a few minutes. No account needed.
- Answer a few questions in a chat
- Download a PDF packet for your bank
If you work in fraud
See which network to act on first, and why.
- Networks ranked by urgency
- A graph of every shared account and phone
- An audit log of each step the agent took
What Rastro will not do
Never freezes money
Rastro ranks, packages and explains. The bank decides what to do with a packet.
Never says anyone is guilty
Briefs that use words like guilty or criminal fail the check and go to a person for review.
Never links by name alone
Two people with similar names stay separate unless an account, phone or link connects them.
Questions
Does Rastro freeze accounts or block payments?
No. Rastro ranks, packages and explains. A person at the bank decides what to do with a packet or a network.
How does it decide which network comes first?
A written rulebook in code scores each network on how recently victims paid, how many victims share one account, and how fast reports are arriving. A SERV model checks that reasoning, and when the two disagree the more cautious result wins.
Can it link two people because their names look alike?
No. Networks form only from shared accounts, phones and links. A similar name appears as a weak, dashed link and never merges two networks.
What if the AI is wrong or unavailable?
Every brief is checked: each claim must point to a real link, and every number must match the data. A failed check sends the network to review. Without a SERV key the engine falls back to plain rules, and the audit log says so.
Is this real victim data?
No. Every name, phone number and account in this demo is synthetic.