27 Jun 2023
In the case of the former, artificial intelligence can greatly increase the speed and effectiveness of decision-making on which assets to lend against and to both improve the quality of lenders’ knowledge and improve the experience of their clients.
I was recently talking to the Head of Real Estate Finance at a specialist property lender, responsible for performing loans, who was beginning to assess the opportunities of AI. The aim of what they wanted to do was pretty simple – at least in theory: to improve the loan book’s overall performance, identify the best borrowers, and decide which are the best properties to lend against.
This requires such them to trawl through vast swathes of data and market data. Successful lenders have always had good market knowledge, but it can require substantial time and resources to find and process – even then, the information can be very variable in quality. This can make all the difference between good lending decisions and uninformed ones – especially in the volatile markets as we have been experiencing over the past 12 months, where speed and accuracy in decision making is vital.
Every month, his company receives a high volume of new applications. The team has to send valuers out to site and then reports have to be written up and decided on. Then the decision needs to be communicated back to applicants. This process of valuing and assessing properties generally takes between two and three weeks. The slower the process, the lower the capacity we have to generate revenue. On top of that, slow and unreliable response times can lead to them having dissatisfied customers. In other words, the company loses out on business to competitors.
An additional risk is failing to react quickly enough to dynamic market changes where the whole market or specific locations become riskier over time due to unforeseen changes. The opposite is also true – would you prefer your collateral to be in a new hotspot where rising values reduce risk over time?
Another challenge is areas being hit differently by over-supply of, for instance, retail and office space because of changing shopping and working patterns. Finally, having some incomplete real estate data means it all needs to be checked manually, causing inevitable delays to the lending process.
So how has the experience been so far for this company? The results from the automated valuations (AVM) have been very encouraging. They added, “Rather than taking weeks to process up-to-date and dynamically changing market valuations, we have the knowledge in minutes”. When valuations are calculated based on a broad range of inputs from private and open public sources, this typically ensures greater accuracy. These valuation sources might be expanded to include transactions data (sold prices), auction results, locality data, land registry and government agencies as well as our own exclusive private sources.
In applying AI-driven market intelligence and automated valuations, the company has removed a lot of the manual work, allowing them to spend more time looking at the quality of their lending and the future strategy for the portfolio.
As a by-product, they have also found that the time consuming process of building reports for the Credit Committee has been significantly reduced, decisions accelerated and faster responses have improved the experience for their customers.
Equally, AI makes it much easier to access and compare the risk to capital. For example, the applications can check portfolio concentrations held by individual investors by seeing their properties in context of their location and risks from external factors like increases in crime rate. These include the local levels of burglary, criminal damage and violent offences. They can also identify changing risks in existing loans. Similarly, they can now model lending risk across different districts, regions, territories, and countries, including the ever-changing potential of extreme weather conditions – such as fire, flood, windstorm, earthquake, and wildfire.
The property data they hold can often be incomplete and incorrect. AI can help to identify ‘what’s not there’ in their knowledge base by helping to spot and fill in any gaps in their data with accurate information.
All this means they can get quicker turnaround times on getting ‘approval in principle’ offers out to new clients without needing any on-site valuers and means they can close new mortgage deals much more quickly. And, most importantly, it allows them to deploy capital lending in the most cost effective and least risky way. All this – and doubtlessly further advances in the future – can greatly improve the performance of the loan book.
