The proposition was launched to the buy-to-let market last year, enabling complex applications to be rapidly assessed through the MPowered Mortgages platform.
The company is now using the same AI to address the residential market at full scale with a suite of products. Its mortgages will be available initially via limited distribution, through intermediaries including Mortgage Advice Bureau, L&C Mortgages, Charles Cameron and Associates, JLM Mortgage Services and The Financial Planning Group.
The current turnaround on an application is 36 hours while cases are being paralleled and checked by human counterparts, but it aims to be able to fully underwrite complex residential cases in real time, once the training wheels are off by year-end.
MPowered Mortgages offers rates starting from 2.24 per cent, with a choice of two and five-year fixed rates for purchases and remortgages. Its larger loan range, aimed at mortgages greater than £500,000, has rates on the five-year terms starting from 2.29 per cent. All products across its prime residential range come with a free valuation.
Stuart Cheetham (pictured), the company’s CEO and co-founder told Mortgage Solutions: “Later this year our ‘submit’ button will become a ‘get offer’ button, with a contract ready there and then. Price and products will be as good or better than the top six.
“This year we’ll be able to fully deliver at point of advice, providing an immediate outcome with certainty during what is the most emotional and stressful financial decision a client will ever make in their lives.
“Any lender can put their products on the platform, then we put that to the broker who then chooses the best outcome for their customer – cheapest, largest, most portable.”
The MPowered platform currently white labels buy-to-let deals from mainstream lenders.
How it works
MPowered has built its AI from scratch but works in a similar way to a price comparison website for financial products like insurance.
The AI can recognise, re-orientate and separate documents even when they’re on the same page. It then extracts, security checks, and verifies the information, and puts together a fully underwritten offer in real time, matching appropriate products to the individual.
It then presents a series of options for brokers to put to their client, putting the final decision in the intermediary’s hands.
The tool is fully integratable with lender systems too.
Richard Fitch, COO, who has been working in machine learning and AI since the 90s, said: “We saw a complex data-rich environment in the mortgage origination process, probably more so than any other field bar medical. There’s the opportunity to apply machine learning and AI.
“AI isn’t used particularly well in any industry as it’s hard to integrate with production systems. But that’s what we wanted to do – create a platform where every process can have AI and machine learning plugged into it at every stage to make it more efficient.”
The platform’s AI learned to process residential complex cases from its inception to release in two weeks.
“The speed at which it can process is only really limited by the central processing unit (CPU) capacity, but it can currently classify six documents in less than three seconds. It’s quite an achievement.” Fitch said.
However, there are limits to the technology and the subjective human touch remains king, particularly when it comes to risk. All documents are currently double checked manually, and then externally quality assured.
Fitch added: “We still have to make sure the documents are decent quality to limit exposure to fraud. Where that fails, by crossing a threshold, it goes to a manual queue to be looked at by a person. You’ll never get to 100 per cent as individuals are annoyingly individual – things will always need a human touch but that’s based on a risk-based decision.
“However, this process will only get faster as the AI learns. Most of it’s already working in seconds and we’re gradually switching off the parallel run processes to enable the system. We’re incrementally testing and changing things and as we get more confident, we’ll use the tech more.”
Cheetham said: “Other companies individually assess consumer circumstances, we choose to automate it and then check it with people. What is defined as complex today won’t be in two years.”
A UK fintech first
MPowered is unique to the British market, citing its closest competitor as Better Mortgages in the USA. Over the past year, the company received relevant permissions from the Financial Conduct Authority (FCA) and won investment backing Barclays, M&G Investments and Citigroup’s SPRINT arm, however investment amounts are undisclosed. The company has since been road-testing its technology.
MPowered has no balance sheet, but its products and services are immediately commercially available as it partners with banks that allow it to use their lending sheets to help originate assets.
Cheetham said: “The tech’s been live in the residential market for six weeks now. We have done it. We can demonstrate it and this is a key part of the change we’re bringing.
“We will rapidly expand this year. Come next year we’ll be the biggest non-bank lender in the market in residential, buy to let, and niche products.”
MPowered is also in talks with around 50 lenders about a range of other potential services the AI can offer.
The company’s long-term goal is to work with all existing mortgage advisers and lenders under the MPowered Mortgages brand or through a direct technology partnership under the MQube brand.
Cheetham said: “There are more products to come later this year and we intend to play in all market segments, giving brokers a one-stop-shop.”
