BMPS 2026: Mortgage sourcing will look ‘undoubtedly’ different in a few years

During a boardroom discussion at the British Mortgage and Protection Senate, Neil Wyatt, chief commercial officer at Mortgage Brain, said the way mortgage sourcing worked today would “undoubtedly” look different in three years and altered by AI. 

However, while the sector is still in the adoption phase, Wyatt said providers should focus on making today’s broker more efficient. 

Mortgage Brain is developing a number of solutions using its own internal AI model, which are scheduled to launch during 2027. This will potentially include utilising an in-house large language model to transcribe a user’s speech to text to search for products. 

Zahid Bilgrami, CEO of Mortgage Brain, said the AI used was owned and had been built by the fintech: “We’ve refined it, we’ve trained it, it sits on our infrastructure, nothing goes anywhere else, the IP [intellectual property] is ours”. 

He said the human in the loop, the adviser, remained essential and AI worked best when inserted into an existing framework. 

Sponsored

Giving brokers the winning edge

Sponsored by Rely

“Embedding AI into the process needs to be such that it enhances your humanness, as opposed to erodes your humanness, otherwise you’re just competing with technology,” Bilgrami said. 

“If you choose to put something between you and your client, what’s the difference between that and a robo-journey?” Bilgrami said, adding that the time freed up by technology should allow advisers to enhance their interactions with clients. 

Wyatt reiterated this, saying the current process involved keying in data and moving around systems to find a suitable product, and Mortgage Brain plans to streamline this. 

Mortgage Brain’s role is to look at the “pain points” felt by advisers, Wyatt said, which could be resolved by technology enhancing mortgage research capabilities and notifying advisers of any information gaps that could support case packaging. 

One adviser said it was key to remember the “grey space” some cases sat in, which did not account for lenders accepting applications at their discretion. 

Wyatt said AI had the ability to return results on a traffic light basis, letting advisers know which cases might fall just outside a lender’s criteria and could be worth a further conversation. 

Probing AI in the right way 

Referring to Mortgage Brain’s AI Charter that was published earlier this year, Bilgrami said some users of the technology did not know the right questions to ask of providers and platforms. 

He said the AI Charter would point people in the right direction. 

Bilgrami raised some concerns about where AI was headed, noting that the current subscription costs to use platforms were being “massively subsidised” by investment capital, as a £20-per-month subscription for the consumer cost between £400 and £600 for the provider. 

“When many people sign up to AI to reduce their costs, in order to be more efficient, to embed AI into their processes, the right question to ask is: how much is that going to cost me in the next 3-5 years?” Bilgrami said. 

He cautioned that some firms made permanent decisions to get rid of staff and restructure their organisation without querying how their cost base might change in the future. 

Bilgrami said it was better to ask technology providers how much a three- or five-year contract would cost, because the token cost would likely inflate over time. He added that the £20-per-month cost was “unsustainable” for providers. 

Users should also ask where the data they put into AI tools goes, as this could be a “class action suit waiting to happen” if people find out their sensitive data has been used to train third-party AI models. 

Further, AI is by nature “probabilistic”, Bilgrami said, meaning it does not give the same outcome to the same prompts. While this was “part of its beauty”, Bilgrami said the downside of this was trying to produce consistent client outcomes. 

He said users should ask technology providers how to guarantee consistent outcomes. 

Further, Bilgrami said many users assumed that when it came to technological advancements, faster meant better, but this should be caveated by accuracy. 

“Where you can afford to take 30 seconds as opposed to three seconds and it costs a tenth of the price, that’s the right discussion to have with whoever you’re speaking to about AI. 

“Having stuff immediately dramatically increases the cost,” Bilgrami said, adding that it was important to determine the right latency of some processes.