Merx founder Andrew Dunning says the applications are being polished by AI and it is not until the lender has spent time digging through the information that major inconsistencies are found.
“The AI tools essentially enable poorly structured loans to look better than they actually are – until the issues are discovered. It is draining resources and nobody is better off.”
He says some lending applications can looking compelling and “tidy” but have clearly not been checked for accuracy, or the sponsor doesn’t know what they are doing.
“It used to be easy to tell, but now we have to wade through pages of good looking rubbish first.
“It is the fastest way to ‘no’ on a funding application.”
Dunning says Merx is not anti-AI, and most of the applications it sees have been done well, but any loan application needs to be read entirely by the adviser submitting it to check for any inaccuracies and to make sure the story is consistent. “It protects their credibility.”
However, Merx, a business and property private credit lender providing finance up to $4.5 million, is seeing a couple of sloppy AI-produced lending applications a week. Although they look polished, Dunning says they were “rubbish” before they went through AI and they are still “rubbish” when they come out, whether they are submitted through an adviser or the applicant themselves.
“The fundamental mistake is not checking the application before sending it. Not articulating to the borrower where the gaps or the inconsistencies in their information are. These should be picked up before the application is sent to any lender. If an adviser’s name is on it they need to do a thoughtful and thorough check for accuracy.”
Dunning says it used to be quick to tell if an application didn’t meet the company’s lending criteria because the mistakes weren’t buried in three pages of a nice looking polished AI summary. “It was maybe a one-pager that had some glaring holes in it so we could make a quick lending decision. Now it takes us longer to reach the same conclusion.”
If AI is used well, it makes the job more efficient, he says. “The gap is around people's fluency, capability, and their professionalism when they're interfacing with lenders on clients' behalf.”
Dunning says the temptation is to run everything through AI as it gives what appears to be a credible response, but if an adviser or borrower doesn’t know what they are looking at it is no good.
As an example, Dunning says a recent business acquisition application had clearly been thrown through AI.
“The applicant put his own financials into the system and then tagged in an information memorandum for the purchase of the new business. His numbers didn’t line up. It was a mess. The application looked compelling but the accuracy wasn’t there.”
He has a tip for advisers and borrowers to protect their credibility and get better outcomes from AI.
“Something like this works in Claude or ChatGPT: Here is my [loan application, feasibility, sales schedule and last six months of bank data]. Check every number in the application against the source documents. List anything that doesn’t match, anything that looks estimated, and the three questions a lender is most likely to ask. Don’t rewrite anything.
“It’s quick, and it finds the gaps before I do.
“Then read the whole thing yourself before it comes to me. If something doesn’t make sense to you, it won’t make sense to me either.”
In situations where there are obvious mistakes Merx has degrees of lending strikeouts. “Some lenders are score-card driven – if you fit the boxes it works. If you don’t then it is a decline. Our approach is a bit more bespoke.
“We work off relationships and try to understand what is happening. It really comes down to the nature of the mistake, but advisers or borrowers start a bit behind if their applications come through with gaps and inconsistencies.”
Dunning says if an adviser sends through a summary that has an inconsistency, it is flagged and means Merx will look more conservatively at the rest of the application. If the rest of the package is okay, a question is asked to get clarification and once that is answered a decision is made.
“We go through verifications and back them up with evidence. We trust, but we verify. If an adviser has not checked the application thoroughly they will get found out eventually. If it is a simple mistake, we can look past it, but if there is a pattern of things made up and not accurate the adviser will get a decline.”
Dunning says if Merx knows an adviser is sloppy they are not going to have a good chance of getting a loan approved in the future. "The files we can trust on first read get a quicker answer.”
Comments
No comments yet.
Sign In to add your comment