How Habito brought digital mortgage fraud to light
Habito is an online mortgage broker and lender, whose mission is to make it easier to get a mortgage and buy a home. It regularly tops lists of best mortgage brokers in the UK thanks to its network of over 90 lending partners, through which it has delivered over $9 billion worth of loans. With its 2020 closure of a $42.6 million series C investment round, it is even expanding its services into its own line of mortgage offerings.
Given this spectacular growth, keeping tight controls to prevent fraud from reaching partners without adding undue friction to customer journeys is a key objective of the company’s onboarding and risk mitigation strategies.
Resistant AI has helped us to drastically reduce both the time it takes to catch fraud, and the amount of fraud that makes it past us to lenders.
Habito’s Fraud Challenge
Broadens your horizons to the risks that exist with digital documentation which are hard to spot with the human eye alone.
The usual escalation between fraudsters and investigators started to break down when it came to digital documents: forgeries on digital account statements and employment records were becoming forensic and invisible to the human eye.
As a result, first-line risk teams were struggling to assess the authenticity of documents, and escalated cases were taking longer for financial crime investigators to resolve. To confirm the authenticity of documents, the teams were increasingly relying on outreach to the institutions that issued the documents, which slowed down decisions — if they got a response at all.
This new situation threatened to bottleneck growth in an otherwise clockwork operation that prides itself on consistent iteration and execution. To address this rising challenge, Habito turned to Resistant AI’s Document Forensics.
New workflows simplify acceptance and speed up reviews
Resistant AI was integrated as part of the application process and workflow triggers were tied to the various verdicts provided to each document. Applications with “Warning” or “High Risk” documents triggered escalations to the Financial Fraud Investigation team.
If the forensic analysis demonstrated clear attempts at fraudulent manipulation, the application could be declined immediately. The team would then be freed up to log the results and analyze them for fraud trends to build their risk intelligence. In less clear cut cases where deeper investigation was needed, time spent on each case was cut by 52 minutes.
Meanwhile, confidence in Resistant AI’s verdicts of authenticity meant applications with documents that met that standard could automatically move on to underwriting stages. In these cases, document assessment was reduced to mere seconds.
Since it only takes 50-80 authentic samples to create new “Trusted” models of authenticity, coverage of the document types Habito dealt with increased daily in production from the incoming stream of documents.