Expert AnalysisJan./Feb. 2023 Edition

Tomorrow’s Technology

The views and opinions expressed in this article are those of the author and do not necessarily reflect or represent the views, policy, or position of Planet Home Lending, LLC.

We’re conditioned to expect technology to have all the answers. Don’t have enough leads? Use a lead generator. Can’t follow your customer in enough detail? Ops staff need help staying organized? Buy yet another reminder/organizer/electronic Trapper-Keeper. Buyers need help finding inventory – Zillow, OpenDoor, Redfin. Who you gonna call? Technology!

But is technology the solution? Yes — but not on its own. At Planet Home Lending, we use the human element to support Artificial Intelligence (AI) to safeguard customer data and keep the Planet Family of Companies stable and strong, no matter the market.

How AI Learns

As our industry seeks to monetize AI and algorithms, it’s critical that we apply lessons from prior AI and algorithms that looked for patterns — from facial recognition to emotion detection — but did not consider ethnicity or cultural differences.

We have the technical skills to teach a machine any repetitive task — when you see X, do Y. As long as we can clearly define what ‘X’ is, the machine has absolutely no problem executing the ‘Y’ over and over, until the end of infinity.

The incongruity happens when the definition of ‘X’ becomes blurred. Let’s imagine that ‘X’ is recognizing that there’s a dog in the picture and ‘Y’ is to make a barking sound. For the machine to recognize a dog, we show it hundreds of thousands of pictures of dogs — big dogs, small dogs, happy dogs, sad dogs, dogs in costumes, running dogs, sleeping dogs, dogs, dogs, and more dogs.

First, we mark where the dog is in the picture. Then we ask the machine to tell us where the dog is, and we correct it if it’s wrong. Eventually, we get to a point (maybe terabytes of pictures later) where the machine is guessing the presence of a dog with almost 100% accuracy. Through the process of consuming this data, the machine gets smarter and smarter. We don’t know exactly what the pattern is, but the machine learns the task through repetition and correction. The larger the dataset given, the more corrections we make, the more accurate the machine becomes at spotting the dog.

Data Makes the Difference

And therein lies the biggest issue. The machine’s “intelligence” is closely tied to the quality of data that it is exposed to. Who defines the data? Who collects the data? Who corrects the machine? How do you correct the machine? All of these are human tasks because we define the success of the machine. We are the creators of the zone of incongruity.

The concept of data is very similar to Lego blocks. You can build different things and have different perspectives with the same blocks. The final product is defined by the plan of the people building with these blocks. Our human touch, our “it” factor differentiates us from everyone else building with the same data blocks. The next big leap (until we get to the strong AI needed for sentient robots) is “us.”

How do we use our experiences, our research, and the yottabytes of data available to us to understand where the market is heading? To better serve our customers, our community and our investors? To stay profitable despite inflation, rising interest rates, and softening home prices?

Planet’s Hybrid Approach

AI bolstered by the doers is Planet’s superpower. As we collect an ever-increasing volume of data, our diverse team is interpreting it through the lens of industry experience. For example, having worked through more than a few periods of intense market inflections, our leadership shifted direction in 2021 because it anticipated what would happen when the Fed raised rates.

Those shifts included:

  • Diversifying our portfolio beyond the average homebuyer to include renovation, state bond, manufactured housing and other niche home loan products
  • Retaining our existing customer base through strong Retail Retention marketing powered by analytics
  • Turning our retail focus to purchase offerings before the refinance boom ended, including launching bridge loans, Cash 4 Home, and buydowns for home sellers

At the same time, Planet’s tech team continued to apply AI and machine learning to support those strategic changes:

  • Ensuring that documents from purchased correspondent loans are indexed and available in our document repository and the loan is onboarded onto our Servicing system within hours of purchase. Our intelligent process tracks every file and uses a rule engine to prioritize which (of the 1 million files plus that we upload every month) needs to be queued next. The process also tracks frequency by business channel and transmittal times per server allowing us to load balance as needed.

Our processes complete the tasks defined within the parameters we set and then provide the data needed for our teams to refine the parameters as needed. They are as successful as the people fine-tuning them.

  • Reducing turnaround times by taking advantage of API connectivity wherever possible — for example, with Fannie Mae to ensure loan onboarding data is available near real-time from purchase, and with our document vendor to maintain a throughput of over 1 million indexed documents (from our LOS, print vendors, prior servicers) uploaded in stacking order
  • Streamlining data availability within the LOS to reduce operational turn-times. We track form load times within our LOS and visits per form, match those to our standard operating procedure to see how best we can arrange data for easier access as well as which forms will give us the highest ROI for a technology rewrite. Our technology allows us to view the stats from multiple perspectives to help make these determinations.
  • Automating fees, including processing, underwriting and tax service, adhering to legal restrictions, and to reduce the need to cure manual errors.

The results of those strategic and technological changes include a larger servicing portfolio increased correspondent originations and strengthened retail divisions — providing the Planet Family of Companies a strong base to withstand the current market.