The financial group Radius, a residential mortgage company, was created in 1999 by co-founders Sarah Valentini and Keith Polaski. Both were veterans of the mortgage industry; Valentini’s focus was on marketing and selling loans, while Polaski’s primary orientation was operations. Headquartered south of Boston in Norwell, Massachusetts, radius began serving residential customers in Massachusetts, but grew rapidly and expanded to 19 states, primarily on the east coast. It also expanded into the sale of property and casualty insurance.
In mid-2021, the radio employed about 200 people and continued to grow during the COVID-19 pandemic. Compared to the Mortgage Bankers Association average, radius had about 30% more revenue from loan production compared to the rest of the industry and about 50% more net revenue than the industry average. It achieved this performance through a combination of innovative and intelligent technologies and highly trained and motivated people. He automated many aspects of mortgage production – or “loan making”, as Polaski referred to him – and used technology to closely monitor the performance of his processes and people. However, it was also named one of the “Best Mortgage Companies to Work For” in 2021 by National Mortgage Newsand one of the Boston Globe’s “Best Places to Work” by the Boston Globe in 2017, 2018 and 2019.
AI and automation in the radius
Motivated by a desire to improve mortgage operations, Keith Polaski and chief technology officer David O’Connor began investigating artificial intelligence (AI) and automation options in 2016. The company had already implemented a supplier loan origination system ( LOS) and a customer relationship management (CRM) system to keep track of leads and customers. But the two executives believed they could do more to automate the loan-making process. The overall goal was for human workers in the radius to touch as little as possible on a client application. I became a consultant to the company to learn more about what they did with this approach.
The search for technology focused on two different approaches. One was to use robotic process automation (RPA) to move the mortgage process from task to task and perform activities such as automatically writing emails to the parties involved with the mortgage application. For this category, radius ended up choosing Uipath as its supplier.
The other was to use AI to extract important information from mortgage documents. Radius found a company north of Boston called AI Foundry that was working on exactly the problem it hoped to solve in mortgage processing. AI Foundry was using image detection technology to identify, categorize and extract the many important data elements from up to 200 different types of forms and even more variations between them. Banks, for example, have many different ways of displaying their statements.
Radius became AI Foundry’s first customer for mortgage banking services, starting with him in June 2018. Polaski said that AI Foundry’s tools initially required a good setup and could only handle about 100 different forms. But the number has increased over time and is now in line with the company’s needs. AI Foundry’s software also performs some mortgage underwriting tasks, but radius uses only a few of them. The company has its own underwriting criteria, makes use of tools provided by US government loan buyers Fannie Mae and Freddie Mac and prides itself on its flexibility in dealing with customers’ financial situation.
Tom Brennan, operations manager at radius, said that when he came to the company in 2014, people were still distributing paper application files around the office. Now, he said, “AI Foundry recognizes 94% of the documents we receive. This removed many human touches from our Disclosure group. Now almost everything moves electronically. The AI system communicates with RPA, which communicates with LOS, etc. It is a complex system, but our people have online work aids for virtually all important tasks. “
Polaski is confident that the AI-based approach is already bearing fruit. Its two largest working groups are loan processors / analysts and official loan assistants, and data on the tasks they work on suggests that 30-35% of their workday a few years ago was spent on indexing and classifying documents. . After two years of investing in the AI system, they know how many person hours they have saved. Overall, they reduced the cost of manufacturing their loans by 70%.
Polaski and his management team refer to AI and RPA resources as their “digital workforce”. He works hand in hand with the human workforce and is positioned as focused on menial tasks. Employees are not concerned about job losses – lightning is growing and hiring – and they often suggest new tasks for RPA robots to perform. AI and automation also helped the radius to overcome the pandemic. Polaski commented: “Last year, we had a 2.5-fold increase in volume. But we are dealing with all of this with less FTEs than a year ago. It has been very difficult to hire in a pandemic, but automation gives us the flexibility to scale our businesses. “
Pavels Danilov, the mortgage launcher
Tom Brennan commented that the “quarterback” job of the loan making process is the Mortgage Processor, who coordinates all stages of the process. Pavels Danilov is the processing manager and, until recently, also processed himself. He’s been in the radius for about six years, so he saw the mortgage process in detail before and after the automation.
At the same time as radius was investigating AI and automation, it began to map the workflows involved in processing a mortgage. The idea was that, for the process to be efficient and automated, everyone would need to perform the same tasks in the same way. Danilov was tasked with mapping workflows – first on a whiteboard and then in software. Both workflows and assistants are now on the Internet for employees to check and learn from, and many of the workflows are built into the RPA system.
In addition to the recognition of the document by the AI Foundry software, the RPA bots took on many different tasks. For example, when the document file arrives, human workers often have to request an assessment, a title review with the closing attorney, a flood zone certificate, and so on. Now all this done by the bot RPA, which composes the emails and directs them to the document file when they arrive.
Danilov said that automation has been a big help to the productivity of his processors. A few years ago, its processors would feel overwhelmed when managing 30 ongoing loans. Now, an experienced processor is comfortable carrying 45 to 50 loans without feeling overwhelmed. This is a time when mortgage guidelines have become more difficult due to COVID’s additional risk measures.
Danilov feels that the “digital workforce” needs to be managed in the same way as the human workforce. Occasionally, he said, the AI system labels a document as unknown, and a human being has to look at and label it manually. It used to happen a lot, but it rarely happens now. Radius is also constantly adjusting its workflows and bots need to be reprogrammed when the work process changes. And whenever there is a new version of the systems with which the bots interact – for example, the loan origination system – they will also have to be reprogrammed. Danilov does not reprogram himself; he turns to Cassandra Pizarro, leader of the software development team at radius, whom Danilov refers to as “the mother of bots”.
Pavels Danilov said that, overall, automation has made the job of the mortgage processor easier. Classifying documents was not a very interesting task and now there is more time to interact with customers, usually to request more information. Communications with customers – usually by email – are still manual, but the system makes up standard emails that the processor can customize. Even that, he said, is a major gain in efficiency.
Danilov also commented on the job measurement aspect. According to him, there are constant measures of how each processor is doing at any time during the workday. Each processor has its own online control panel reporting its own performance and, as Processing Manager, Danilov has performance indicators for each processor’s work performance. There are standards for each task and the panel lets processors know if they are working according to the standard. “If you’re substandard, it adds stress to the job, but it’s also a motivating factor,” he said.
Overall, he summed it up: “People like to work in lightning. We had a rally yesterday. It was virtual, but people were in a good mood. The question, ‘Why do you love lightning?’ Was asked of several people. Most said it was the other people who worked for the company. But a loan officer mentioned technology and automation in the radius. ”Although the digital workforce may not inspire as much love as the human, it seems to be an essential element of the company and its success.