Brokers that do not use artificial intelligence (AI) will eventually be replaced by those that do, according to Mortgage Advice Bureau (MAB) innovation director Matt Lowndes.

Speaking at an AI session at the Mortgage Strategy MIT Live event in London, Lowndes predicted that AI would become increasingly useful for brokers, and that customers would be likely to expect it from their adviser.

“I do believe that brokers that don’t use AI will be replaced by brokers that do use AI,” he stated.

AI isn’t here to replace mortgage advisers. It’s here to accelerate the way we work

Lowndes added that one third of adults already made use of AI in some form to manage their money every week.

“That change is happening,” he said. “It’s here, it is real now.”

However, Lowndes said MAB brokers were finding real benefits from using AI to supplement their work. For example, broker accuracy with documentation was around 80% without AI but rose to 99% when the technology was used to help.

Lowndes added that AI also helped with the speed of checking documents. Tasks that would take a human 45 to 60 minutes, such as checking a client’s bank statements, could be completed in just 60 seconds by using AI, he added.

“The human is still in the loop,” he said. “AI is not just a bolt-on; that’s really important. But it’s also really important that the human is still smack bang in the centre of it.”

If you are going to put customer information into AI, you have to make sure you have the right system with the right agreement to do that

Likewise, MAB has seen a 65% improvement in case hand-ling and a 164% increase in complaint closures when using AI.

The broker’s analytics software shows that 90% of MAB customers upload mortgage documents from their mobile phone, Lowndes went on. The same software reveals that customers send documents to MAB at all hours of the day and night.

In part to respond to this, the broker has launched an AI chatbot called Morty. This chatbot can answer customer queries at any time, even when human advisers are sleeping.

“It’s very good at answering questions,” said Lowndes.

Speaking at the same event, Connect Mortgages chief executive Liz Syms said there were many ways in which this technology could be used to help the mortgage market, such as by building websites and software, editing photos and translating speech into different languages.

While a lot of AI can provide a productivity boost, it doesn’t provide real, meaningful transformation

“AI isn’t here to replace mortgage advisers,” she said. “It’s here to accelerate the way we work.”

Connect had launched mortgagelab.ai this year, Syms said, which gives brokers advice and guides on how to best use AI to help their business.

The current version of mortgagelab.ai will be expanded in the new year, Syms added.

Lendwell co-founder Jamie Lawless said there were commonly held fears that AI would replace the role of mortgage broker, but this did not have to be the case.

“AI should make us more human, not less,” he said. “While a lot of AI can provide a productivity boost, it doesn’t provide real, meaningful transformation.”

Lawless cited “three big problems” with AI in the mortgage market, which must be navigated: AI does not understand lender-specific context; it does not explain its reasoning; and it does not plug into adviser workflows.

Our chatbot is very good at answering questions

“To use AI to transform your business, you need one that understands context,” Lawless continued. “To be effective, AI must tackle this.”

He was confident about the future role of brokers.

“I think we will need more human advisers, not fewer, in this industry. AI can remove the work that never needed to be human in the first place. The better the AI, the more human the advisers can be.”

In the Q&A section of the discussion, the panel talked further about some of the limitations of AI within the mortgage industry, as well as issues to be aware of.

Lowndes said that MAB had deliberately put restrictions on the firm’s use of AI, to keep the technology’s frequent errors and ‘hallucinations’ in check.

AI models are known to generate some responses to questions that sound certain and plausible but in reality are wrong, and sometimes appear to be invented.

AI is not just a bolt-on. But it’s really important that the human is still smack bang in the centre of it

“We know it makes stuff up,” added Lowndes. “We have gone small on purpose. You have to be careful with this stuff.”

Syms also added a note of caution, saying: “If you think about AI as an employee, even if it gets things right more than many human employees do, you will still need to check what that AI employee has done.

“If you are going to put customer information into AI, you have to make sure you have the right system with the right agreement to do that.”