by Ari Gross, CEO & Founder
Not long before I wrote this blog, President Biden signed an executive order seeking to balance the promise and risks of artificial intelligence (AI). Around the same time, over in the United Kingdom, a diplomatic breakthrough was hailed at a summit that achieved an unprecedented international declaration on AI safety.
Action by legislators and regulators is a sign of the growing prevalence and significance that AI has for our economy and society. Leaders of companies that use (or plan to use) AI should see these interventions as a prompt to improve their understanding of the technology.
The Inner Workings of AI
The mortgage industry has not typically been at the forefront of AI innovation, but some of the latest advances are highly applicable to mortgage processes. A growing number of mortgage firms are beginning to make investments in AI tech in response. So what is true artificial intelligence?
If you’re trying to understand how AI could be usefully deployed in your mortgage business, it can be quite a challenge. The origins of AI reach back to the middle of the last century. Waves of progress have produced evermore practical and valuable applications, but the long history of AI development means that some of the systems we use today can be extremely difficult to explain, especially to non-technical audiences.
As a result, few vendors of AI-based solution explain the inner workings of their technology. This may be due to commercial sensitivities or the fact that certain aspects of AI are inherently unexplainable. However, mostly it’s because it’s easier to market AI on its results – plus the current media hype can be quite motivating!
For many technology buyers, a proof of concept that apparently confirms promises of system performance is seen as adequate. Shocking as it may seem, not all AI vendors are completely transparent about how they deliver on their claims! I would urge you to probe deeper.
I’ve studied and worked in the field for more than 40 years, gaining several patents in areas related to imaging, machine learning and automation. This has given me a firsthand appreciation of the gap in understanding between those building AI systems and those using them. This is a risk when AI makes or contributes to critical mortgage industry business decisions, from offering or denying loans to the financing provided by the secondary market. As consumer rights are strengthened, it’s likely that processes that involve mortgage machine learning or AI will be subject to stricter rules and governance requirements.
The Potential of AI in the Mortgage Industry
The AI we’ve built at TRUE is designed specifically for the mortgage industry. Fundamentally, it solves the thorny problem of turning the messy information in borrower documents into structured data that is reliably accurate and can be fully trusted.
And once all this data is made sense of, it’s used to power many of the automated solutions that our industry has bet so heavily on, such as POS and LOS platforms. Until recently, these are investments that have not fully delivered on their promise of automation as the data needed to fuel their processes has been so unreliable. Undoubtedly, AI is becoming the missing link in our industry’s long-held desire to automate.
But the automation doesn’t have to stop with data. There are new AI technologies, such as Generative AI, that can help us pull insights from this data and enhance how we understand all this information.
Imagine a scenario where AI is used to process, structure, validate and store important information automatically. Then the AI analyzes this same information and generates insights, nudges and suggestions – such as an income analysis process, where the generative AI model outputs, in human language, a summary of the borrower’s income.
This kind of innovation would mean that instead of an underwriter taking the time to review all the details of the analysis, the summary could tell the whole income story. This would save considerable time and effort – while reducing time to decision and risk. This same process would play out over and over. From income, to fraud, to source of funds and beyond, generative AI could help lenders understand and see the whole loan story, in human terms, without having to consume valuable time looking through reports, graphs, tables, and many other sources to come to a decision.
The potential is huge, but it’s very important to note that a generative AI model cannot work well if it is not fed correct data. That’s why, before embarking on experiments to explore the potential of AI, it’s vital for mortgage businesses to build on their foundational tech stack with accurate data classification solutions and loan document indexing and extraction software.
All of this has obvious benefits for efficient mortgage manufacturing and reduced time-to-close, but the potential to expand into other use cases significantly widens the scope for ROI too. I believe that a better insight into what’s going on under the hood will help prospects with their evaluation. It increases confidence in what our AI can do, helps to mitigate concerns about risk, and shows how to get the most from an investment into such technology.
How TRUE Can Help
I see it as my role, as a business leader and technologist, to explain not only the results that our solutions deliver but the principles of how they are achieved. I’ve written a paper that explains our AI. It takes you through the business problem of processing borrower documents, why AI can offer a reliable solution, and describes the principles of how our AI actually works.
If you’re interested in learning more about this, watch our on-demand virtual discussion with Bob Noble, our Chief Product & Technology Officer and I, titled: A Simple Way to Understand Lending AI.
Watch the webinar recording >>
And there’s more. We also created a tool that gives those of you actively evaluating solutions the means to test our AI without having to speak to a salesperson or take a demo – you can do that when you’re ready. Our Try Now tool lets you upload your own examples of borrower documents and see the outcomes of mortgage document classification and data extraction in moments.
We wanted to offer self-serve evidence of our AI because some solutions on the market blend AI with human intervention. That’s not inherently a bad thing, but it isn’t genuine AI and introduces time delay and risks to data security. With TRUE, you’ll see results so fast that you’ll know that it really is our AI at work.
I hope you find my paper and the Try Now tool to be helpful. If you have questions, I’d be happy to answer them personally.
