One of AI’s greatest successes has been the transition from manual to automated document processing, which has freed up laborers to concentrate on providing genuine value by eliminating countless hours of tedious, error-prone manual work. This accomplishment has been greatly aided by Google Cloud’s Document AI.
However, new methods of document intake may expose businesses to varying degrees of danger and open up fresh chances for fraud, especially in high-risk sectors like the financial services. Regulations pertaining to Know Your Customer (KYC) and anti-money laundering (AML) demand that these institutions gather and process records in order to better understand their consumers. Without sufficient controls, automating certain procedures might actually make them more risky.
Since implementing operational efficiency like automation into their organizations is often a core organizational aim, the dangers for FinTechs are particularly significant. However, this also presents opportunity for many forms of malicious actions.
A excellent illustration is when organized criminals establish phony accounts that may be used for fraud and money laundering. Due to this, one of the main difficulties faced by FinTech Product and Operational teams, especially in the automation of client onboarding, is fraud resilience.
Due to the fact that most digital document forgeries leave very little to no apparent evidence, manual verification is unable to detect them. After analyzing tens of millions of documents, Resistant AI discovered that 15% of company registration certificates submitted globally when opening corporate bank accounts are fake, and that up to 17% of digital bank statements used for loan applications, account openings, and other purposes have been falsified.
An inexpensive approach to expand assaults beyond human controls and build legions of disposable accounts that may be exploited as fictitious money mules is provided by serial fraud attacks that depend on simple-to-find document templates.
To safeguard any workflow created using Google Cloud’s Document AI, Resistant AI created Document Forensics, an easy-to-deploy API that can automate document fraud checks.
Automated checks for document fraud
A rapid judgment to accept, reject, or elevate a digital document for further assessment is the main objective of Resistant AI’s Document Forensics. Whether in PDF or picture format, the supplied document is qualified, categorised, and confirmed in a matter of seconds to speed up the review process and offer a smooth and simple experience.
- Readability: Is the document readable? In this phase, the papers are examined to make sure that they are of a caliber that allows them to be processed for optical character recognition (OCR) and that no fraud is being concealed by camera flashes, poor resolution, blurry images, or other issues. Customers may utilize rejection reasons as feedback when submitting a new version.
- Classification: Is the document the proper type? This procedure makes sure that you only accept papers that satisfy your acceptance standards. If all you’re after is utility bills, you shouldn’t be accepting and processing ID cards.
- Verification: Is it reliable? This process examines the document 500 times for evidence of manipulation. In order to detect patterns of document reuse, template farms, and widespread fraud efforts across accounts, it will compare all supplied papers after doing individual analysis, taking into consideration everything from metadata to the color temperature and scene composition.
Together, these three processes ensure that human interaction isn’t spent on instances of obvious fraud that should be rejected or on documents that should be automatically handled by Google Cloud’s Document AI extraction since they don’t seem to have been tampered with. A user interface (UI) that makes every finding that led to the judgment visible to the reviewer when a document is escalated for review helps the reviewer make decisions faster.
Payoneer, a significant worldwide merchant services provider, was able to cut the number of manual document fraud inspections they conducted from 92% to only 18% of all submissions. Additionally, they are now preventing hundreds of daily attempts at serial fraud during onboarding.
The top digital mortgage broker in the UK, Habito, was able to shorten fraud investigations by 52 minutes per case and discover 32% more fraud than their previous fraud solutions.
How to integrate Google Cloud Document AI with Resistant AI Document Forensics
Customers may expect to fully use the security and privacy standards they’ve come to expect from Google Cloud to apply to Resistant AI by using the same cloud, which will hasten the procurement process.
In addition to merely collecting data from papers, Document AI also enhances the output using the Google Knowledge Graph to cross-reference business names, addresses, phone numbers, and other facts with online businesses.
[…] automatically. Some chatbots have simple menus, as shown below. However, advanced chatbots can use AI and NLP to understand user input and navigate complex human […]