Loan Origination

Introduction

Loan application fraud in the UK has become an increasing concern for financial institutions and regulatory bodies. This type of fraud occurs when an individual or business deliberately provides false information or conceals relevant details in a loan application to obtain funds they might not otherwise qualify for. According to data from UK Finance, loan fraud losses in the UK increased by 40% in 2022. Over £1.2 billion was stolen through fraud in 2022 – within that figure, unauthorised fraud losses were £726.9 million.

Common Tactics include:

• Misrepresenting Income e.g. the true source of income or its frequency
• Inflating or Misrepresenting Assets e.g. over estimating a value of an asset or changing its description.
• Concealing liabilities e.g. concealing an existing debt
• Stolen or synthetic identities e.g. using someone else's Identity, address, business name or social media profile. Some fraudsters generate synthetic identities and documents to support their loan applications
• "Loan stacking" – applying for multiple loans within a short timeframe to circumvent credit checks – has emerged as a prevalent tactic.

Regulatory authorities, including the Financial Conduct Authority (FCA), continue to emphasise the importance of stringent anti-fraud measures and the role of compliance in minimising these risks. As fraudsters adapt their methods, the UK’s financial sector must remain proactive, balancing robust verification processes with the demand for streamlined, accessible lending.

Example 1 Using Gen AI to detect financial crime in Loan Applications

In this example a UK Food Manufacturer “The Real Bakery Company” has applied for a £30k asset based fiance loan and listed 4 ‘ Nesbitt Walk in Freezers’ as collateral for finance.

As part of the loan underwriting procedures the applicant is asked to provide serial numbers for the freezers.

The Results

The serial numbers you've provided—0423 1280 010077, 0292 1131 000031, 0422 1280 010077, and 9992 8212 23333—appear to follow a format commonly used by certain refrigeration equipment manufacturers. In this format, the first segment typically represents the manufacturing date in a YYMM (year and month) format. For example:

0423 would correspond to April 2023.
0292 suggests February 1992.
0422 indicates April 2022.
9992 is unconventional, as the '99' year prefix combined with '92' month is atypical and may indicate a different coding system or a potential error.

<SerialNumbers>
  <SerialNumber>
    <Number>0423 1280 010077</Number>
    <Year>2023</Year>
    <Month>April</Month>
    <AdditionalInfo>1280 010077</AdditionalInfo>
  </SerialNumber>
  <SerialNumber>
    <Number>0292 1131 000031</Number>
    <Year>1992</Year>
    <Month>February</Month>
    <AdditionalInfo>1131 000031</AdditionalInfo>
  </SerialNumber>
  <SerialNumber>
    <Number>0422 1280 010077</Number>
    <Year>2022</Year>
    <Month>April</Month>
    <AdditionalInfo>1280 010077</AdditionalInfo>
  </SerialNumber>
  <SerialNumber>
    <Number>9992 8212 23333</Number>
    <Year>Unknown</Year>
    <Month>Unknown</Month>
    <AdditionalInfo>8212 23333</AdditionalInfo>
  </SerialNumber>
</SerialNumbers>

The LLM identifies one of the serial numbers 9992 8212 23333 as inaccurate whilst the other 3 appear to conform to the Nesbitt Freeze standard.

The incorrect serial number may perhaps attributed to a user input error. Nevertheless identifying these errors for both customer service and potential future fraudulent activity is important.

LLM has also been trained to also generate accurate current market values for the freezer with referenceable examples for assets.

Real life examples of similar show for sale in the UK for comparison

If you're seeking 2-year-old walk-in freezers available in the UK market, here are some options to consider:

1. CaterQuip

CaterQuip specializes in second-hand commercial kitchen equipment, including walk-in cold rooms and freezers. They offer bespoke projects ranging from small walk-in fridges to large warehouse-style freezers. For instance, a recent listing featured a Foster Walk-in Fridge/Freezer priced at £4,750.00.

2. Secondhand Catering Equipment

This platform offers a variety of used catering equipment, including walk-in freezers. A recent listing included a CellTherm Walk-in Fridge Freezer in good working order, dismantled on 01/08/24, with full safety doors and curtain, Danfos controllers and condensers, and racking, priced at £2,500 + VAT.

The applicant is asked to make a 5 second video of the the serial number plate to validate the asset is real

The Maximum value of the assets is £10,500 and the loan is approved.

Example 2

Delli UK Ltd has a loan application for £5k providing 2 serial numbers for Charlie Charcoal ovens as part of the requested documentation.

Serial numbers for Charlie Charcoal ovens Delli UK provided

LLM identifies the serial number was seen before in the previous loan application for The Real Bakery Company for a different asset (see example 1.

Statistically speaking the odds of a genuine coincidence in serial numbers being the same for both application is 10 billion to 1 one.

If this were indeed a real life scenario, there is a strong possibility that application would either be refused by the FI or escalated for enhanced due diligence.

Conclusion

Generative AI can be a powerful tool in loan origination for detecting financial crimes by identifying complex patterns, anomalies, and inconsistencies in application data that might otherwise go unnoticed. Traditional fraud detection methods rely heavily on rule-based systems that can struggle to keep up with evolving financial crime tactics. In contrast, generative AI models can analyse vast amounts of data from multiple sources, uncovering unusual patterns in borrower profiles, transaction histories, and documentation that may indicate potential fraud or money laundering. By simulating realistic financial scenarios, generative AI can also help train other models to recognize subtle, high-risk behaviours, thereby enhancing the overall accuracy and effectiveness of risk assessment in the loan origination process. This advanced capability helps financial institutions improve compliance, reduce losses, and support a more secure lending environment.