A broker with a taste for technology says AI's potential in the industry is real – but someone will have to pay for it
The parallels between software development and mortgage broking may not be immediately obvious, but broker Stuart Phillips (pictured top) of AALTO Mortgages has been thinking carefully about what the rise of artificial intelligence (AI) means for both professions – and where the similarities break down.
Phillips, who has been described as someone who "knows just enough to be dangerous" when it comes to technology, has experimented extensively with AI coding tools such as Claude Code and ChatGPT. His conclusion is that the two disciplines share a fundamental challenge: taking complex, sometimes unpredictable clients, understanding their objectives and constraints, and producing a solution that works for them.
"In essence they both take complex and sometimes irrational or unpredictable clients, understand their objectives, budget and constraints and work to produce a solution that benefits the client and results in revenue for the business," he said.
The case for AI in mortgage broking rests heavily on data. Lenders publish criteria and rates publicly, creating a large and accessible information base. The difficulty lies in cross-referencing multiple criteria points simultaneously – something Phillips believes AI is well suited to handle.
"When you are trying to find a lender that will accept three or more criteria points that can be extremely complex and time consuming and it's easy to miss opportunities simply because the lenders wording is ambiguous or spread across multiple large documents," he said. "Humans are simply not great at retaining all that data in their minds at once."
However, Phillips is candid about where AI falls short. Both brokers and developers spend the majority of their time not solving the core technical problem, but managing relationships – with clients, solicitors, estate agents and accountants, he said. That interpersonal dimension remains beyond AI's current reach.
The broker also raises a concern about client expectations. Borrowers are increasingly using AI tools to research their options before approaching a broker, which can introduce confusion. "It's frustrating having to explain why terms the AI spits out are not relevant to UK law or practice or are simply made up," he said.
The more fundamental obstacle, in Phillips's view, is cost. He argues that the industry is currently in a "honeymoon period" in which the benefits of frontier AI models are visible to users through products such as Microsoft Copilot and Google Gemini in search results, without the true cost of access being passed on.
"The true cost of API access to frontier level models is significant, around £25 per million tokens which you can easily burn through in an hour or two of discussion," he said. "Scale that up to a broker using this daily and costs could easily be hundreds of pounds a month per user."
That raises a pointed question for any company looking to build broker-focused AI products: would brokers pay a significant monthly premium for efficiency gains in a task that, while important, is not the most time-consuming part of their role?
"AI is very impressive, but for so many use cases, the benefits simply do not justify the real cost that will have to be paid," Phillips said.
His assessment suggests that while the appetite for AI in the mortgage market may be growing, the business case remains unproven – and the companies best placed to deliver such tools have, so far, held back.
Want to be regularly updated with mortgage news and features? Get exclusive interviews, breaking news, and industry events in your inbox – subscribe to our FREE daily newsletter. You can also follow us on Facebook, X (formerly Twitter), and LinkedIn.
