The robotics age has come to the mortgage industry

Register now

A growing number of lenders, servicers and vendors are implementing robotic process automation to make the mortgage process more time- and cost-efficient.

And the need for increased — and improved — automation has perhaps never been greater, as the cost to originate a loan has remained stubbornly high.

The average per-loan cost to originate a mortgage was $6,969 in the third quarter of 2016, and costs have hovered between roughly $6,000 and $8,000 per loan since 2013, according to the Mortgage Bankers Association. Similarly, origination turn-times remain stubbornly high.

"We haven't really broached the 45-day on average close," Marchetti said. "If you look across the industry and you look at Fannie's numbers, they're still at 70 days from origination to delivery to them. That's a lot of time to put the customer through."

Indeed, the average time to close for all loan types came in at 46 days in February, Ellie Mae reported, down from 51 days the month prior.

Speed and efficiency could quickly become all the more important if the market continues its shift toward a more purchase-oriented environment. As of March 15, the Mortgage Bankers Association forecasts that high rates will cut total 1-to-4 family mortgage origination volume to $1.6 trillion in 2017, from $1.89 trillion the year before. That figure is expected to decrease even further in 2018.

But while the origination sector preps for tightening margins, the servicing side of the industry has long felt the squeeze. Between 2008 and the first half of 2016, the fully-load operating for servicing performing loans rose from $59 to $228, according to the MBA. On the nonperforming side, costs per loan jumped from $482 to $2,522.

At the heart of these operational struggles is the increased regulation the mortgage industry has faced. From the implementation of TILA-RESPA Integrated Disclosure rule to the updates to the Home Mortgage Disclosure Act, lenders and servicers have to collect and to process more data than ever before. And much of that data continues to be handled in a manual way.

The TRID rule implemented in October 2015 singlehandedly increased lenders' back-office fulfillment and post-closing costs by $209 per loan, according to the results of a lender survey conducted by Stratmor Group released in March 2016.

Compliance isn't the only pressure lenders are facing though. Industrywide, a major shift is occurring toward a fully digital mortgage process, which necessitates state-of-the-art systems and platform upgrades to handle electronic documents. At the same time, programs like Fannie Mae's Day 1 Certainty are rewarding companies that embrace data-driven processes.

That's where robotic process automation and artificial intelligence technology come into play.

"There is a lot of groundswell and excitement right now around robotic process automation and artificial intelligence," Suzanne Arden Powell, a partner and mortgage originations and servicing solutions executive at IBM Global Processing Services said.

IBM's loan servicing subsidiary Seterus integrates robotic process automation technology into its operations. IBM also serves as a provider of RPA technology for mortgage industry clients.

"We have a lot of clients that have this as the top thing they want to do," Powell said. "In the next 12 months, you're going to see a lot of lenders have a lot of functions use robotic process automation."

More traditional forms of automation occur on the back-end of a system and often within a single system altogether. Robotic process automation, meanwhile, is designed to work across systems. With RPA, software interacts with and interprets existing applications to process transactions, handle documents and data and trigger communication with other systems.

"This is a class of software that exists to operate in multi-application world," said Charles Sutherland, vice president of product management and strategy for lending solutions at Fiserv.

RPA software functions at the graphical user interface level — the level that human users of these systems and platforms see.

"We're creating software robots that mimic what humans do," said Mark Davison, a partner at technology research and advisory firm ISG, which specializes in RPA. "They can sign on to an application, make menu selections, make transactions, copy and paste information. When we build these software robots, we are replicating the current process."

To this end, one way that RPA bots are programmed is by using screen capture technology to record what a human user sees during the task that's being automated.

And because RPA functions on the surface level, and not on the back-end, of a system, it is seen as a more cost- and time-effective form of automation. Traditional automation requires an overhaul of the entire program or software. Because RPA is designed to interact with systems, rather than be intrinsic parts of them, it can be implemented more quickly and cheaply.

"We look at RPA as a bridge to help fulfill the short-term needs of something that can be simplified or made more consistent against more transformative changes," Marchetti said.

For all its benefits, RPA technology can have its drawbacks. While many consider RPA to be adjacent to, or even a subset of, artificial intelligence and machine learning, it differs from that more advanced technology in important ways. RPA is like a parrot — it can replicate how a human does something, but it doesn't necessarily know why.

Consequently, RPA software relies on structure to carry out its tasks — without certain inputs it won't function. Companies must always be monitoring robots to make sure nothing has gone awry, particularly if there is ever a transition to a new system.

"You have to consider your robots as another set of staff," Cheryl DeRoche Johnson, managing director of innovative solutions at MUFG Union Bank, said. "You wouldn't have staff do a job without having a process and procedures in place."

For reprint and licensing requests for this article, click here.
Mortgage technology Analytics Automation Mortgage Bankers Association Ellie Mae Stratmor IBM Fiserv LoanDepot