With more and more lenders turning to Artificial Intelligence (AI) software, lenders and customers alike are discovering that technology and humans both come with their own unique sets of pros and cons.
Using AI software as a way of deciding how lenders can borrow money and who is approved for a loan can help to overcome some of the bias that previously tarnished the lending process’ reputation. An algorithm-based software can help to prevent people from being turned down for a loan due to characteristics such as gender, race and sexual orientation.
AI software isn’t bulletproof when it comes to deciding who can lend money though. The system is often inputted with historical credit data in order for it to make an informed decision. The data that’s submitted is often biased credit data which has been drawn from years of bias in lending markets. It’s crucial therefore, that the data used to feed AI systems is checked thoroughly and regularly.
The UK recently issued lenders guidance on using AI software, along with Singapore and some European countries, that required firms to promote fairness in their use of AI, particularly when using it to inform decisions on lending.
But with AI looking increasingly set to become a mainstay of many industries in the future, the question is, how can lenders make the most of it while still ensuring the process is fair to all? One of the key solutions will be to build an AI system that doesn’t just focus on historical accuracy and matching current decisions to previous ones. The systems will need to be trained and tested to not just make a decision based on how money was lent in the past and instead on how the money should have been lent in a more equal society.
Many lenders are beginning to understand the need to judge lending abilities on a case-by-case basis with borrowers having various external factors that impact their lending abilities.
Online lender Upstart uses an up to date AI system to make key lending decisions, with co-founder and CEO Dave Girouard acknowledging that there is still room for better understanding of how human behaviour and external factors play their part in impacting a lenders repayment abilities.
The US based company’s system uses 1,600 data points in order to determine creditworthiness which include colleges an applicant attended, the degree obtained and their profession. It’s also the only FinTech company to receive a no-action letter from the Consumer Financial Protection Bureau, meaning the firm has the bureau’s approval to pursue AI-based lending as long as Upstart submits its data on loan applicants, approvals and rejections on a regular basis.
Credit modelling is likely to always be a bit of an imperfect science, no matter the advances of AI systems, however more lenders are uncovering the perks of using an AI based system, providing they input and regularly assess how they are using historic data.