The 90-Day Question For Lenders Measuring Their Automation ROI
A recent KPMG survey of Canadian businesses found that 93% of organizations are now using AI in some form, up sharply from the year before. However, only 2% said they’d seen a measurable return on their generative AI investment. That number makes it pretty clear that adoption and implementation have moved rapidly, but the proof hasn’t kept pace with them. It also raises an obvious question for any operator considering a new AI or automation vendor: where does this technology reduce cost or drive revenue, and how will you know within 90 days?
At a time when the majority of mortgage lenders are racing to adopt new tech, especially AI based tech, that question is critical. In fact, it’s the first question decision-makers should be asking before they seriously consider their next vendor’s pitch. An honest answer to that question will come from the answers to a few more. Does the tool operate inside a real-life production workflow rather than simply co-exist with it? Is it trained on data specific to the industry/market it serves? And does it address a problem expensive enough that solving it has a positive impact on one’s P&L? Aa affirmative answer to all three questions is a fair and effective requirement to demand of any potential technology vendor.
These requirements are especially applicable to custom built RPA and AI bots. A bot that operates from inside an LOS, rather than a separate portal someone has to remember to check, passes the first test. A model trained on things such as TRID disclosures and the document types specific to a mortgage transaction such as applications, credit reporting or flood certifications meets the second, since general business language doesn’t often capture the terminology this industry runs on. And functions like employment verification, appraisal review or closing disclosure reconciliation meet the third, since they consume enough staff time every month that automating even part of them shows up in the numbers.
The clearest metric for answering the 90-day question in this industry is hours returned. How many hours of staff time did a process require before a bot or other technology took on part of it, and how many hours does it require after implementation, measured the same way both times? A lender that automates the majority of its income or asset verification process, for example, can compare the resource hours spent before deployment to the hours spent 90 days later and know exactly what changed. That figure comes from the firm’s own operation, independent of anything in a vendor’s sales deck. The same holds true for document indexing, condition clearing or any other function with a defined, repeatable workload. A vendor unable to point to that kind of before-and-after on a function specific to your operation, rather than an average across their entire client base, hasn’t really answered the 90-day question.
Another important question to ask is who owns the data and the workflow logic behind the automation once the relationship ends? A bot’s value comes partly from the rules it’s been given, as well as the documents on which it’s been trained and the way it’s been configured to a firm’s specific process. If none of that can be exported or rebuilt elsewhere, an agency or lender has effectively handed part of its operational memory to a vendor it may not have a relationship with in a few years, whether that’s because the vendor is acquired or simply doesn’t work out. Custom-built automation should provide a client a clear answer to that question before the first bot ever goes live, not after.
The KPMG numbers represent a broader industry metric, but mortgage businesses should be asking each and every potential tech vendor the same two questions: what’s the measurable, 90-day answer, and what happens to our data and our workflow logic if this relationship doesn’t last? Be sure to ask both questions and demand specific, concrete answers, instead of general reassurances, long before signing any contract.
Jimmy Lewis is the CEO and Co-Founder of TrueFocus Automation, a Plano, Texas-based provider of custom software bot development and Automation as a Service solutions to the title and mortgage industry. He can be reached at jimmy@truefocusnow.com.
