Artificial intelligence (AI) is now used to execute searches on Google and Bing, verify credit card transactions, make product selections on Amazon and suggest movies on Netflix. As you go about your daily life, consider areas where you see AI applications. By identifying these examples, you will gain an understanding of AI’s potential.
As a seasoned mortgage professional, you are well-versed in the language of lending, but with AI concepts gaining interest in the mortgage industry, it is crucial to add the latest AI buzzwords to your vocabulary. These terms — from machine learning to large language models to chatbots — will have a practical effect on your company.
The McKinsey Global Institute estimates that generative AI, or GenAI, could add between $200 billion-$340 billion in value annually across the global banking sector — mostly from improved productivity. The value of GenAI equates to 9%-15% of operating profits.
A report by KPMG last year indicated that 61% of banking technology leaders believe some form of AI will be critical for the business to achieve its short-term goals. The report suggests technology leaders are being inundated with business requests for GenAI support.
Fundamental change
Two years ago, OpenAI caused quite a stir with its launch of ChatGPT — an advanced conversational GenAI model designed to understand and generate human-like text. Though GenAI is in the formative stages, it is the missing link required to drive significant change since it makes it easier to build AI and less expensive to implement.
At a macro level, GenAI’s capabilities to drive impact are likened to the introduction of iPhone in terms of its transformative potential. A recent study from Penn and OpenAI concluded that around 80% of the U.S. workforce could see at least 10% of their tasks affected by AI. Both GenAI and the iPhone represent paradigm shifts — anybody can see the impact of smartphones firsthand, while GenAI is emerging as a methodology to change the industry, from operations to borrower engagement to delivery. All these combined have the potential to fundamentally change the mortgage industry.
GenAI can be leveraged to create documents and communications, such as letters, emails and images. It can be used to translate communications into multiple languages or even to generate programs by converting natural language into code. GenAI can make decisions that are nonlinear; with GenAI, decisions are not limited to algorithmic, linear logic.
GenAI has been effective at fraud detection as large amounts of data must be reviewed. Traditional fraud detection methods can miss complex patterns with rule-based systems, but generative models more effectively learn normal behaviors then detect deviations from this behavior to flag outliers for further investigation. Since fraud occurs infrequently, identifying subtle patterns of change and doing so timely helps to suspend suspicious activities early in the process.
GenAI can be leveraged in loan originations to streamline the process of collecting client information, which is a laborious task. In a Moody’s Analytics survey, 56% of bankers responded the biggest challenge related to the loan process was the manual collection of data and subsequent back and forth with the client.
GenAI retrieves and processes loan origination information. It does this by automating repetitive tasks, enabling significant efficiencies. Using GenAI, documents are validated, data extracted and feedback provided based on specific loan programs to automate workflow processes, like asking applicants for relevant details and documents.
Realize efficiencies
AI can be leveraged to improve accuracy in predicting income levels based on data available. AI models perform historical data analysis to identify high-risk scenarios that may not be evident in manual reviews, but that are important to avoid buyback risk.
In addition to time savings, errors due to manual processing are reduced and resources are deployed on higher value tasks, increasing job satisfaction. For the mortgage lender, reduced time to approve the loan can increase pull-through rates and decrease fallout, while allowing the front office to focus on lead generation (which AI can help with as well).
GenAI is used to accelerate software development to realize business results faster. Programmers, for example, are reporting being able to produce material portions of code with GenAI; these programmers use GenAI to generate code from public sources and then customize it and add code to fill in the gaps. Companies have been leveraging GenAI to automate regression testing to verify code, reducing production errors.
Everyone has had experience with automated call centers providing generic responses that fail to answer the key question. Enter GenAI with technology that gets smarter by understanding what types of responses are effective. Analytics are applied automatically to determine the best responses. This frees people to do more complex tasks.
The GenAI technology learns from large learning models (LLM) — think of these models like having an encyclopedia, a dictionary or large amounts of key data available to learn from on the fly. The LLMs using existing industry data are coupled with proprietary data sources to provide highly customized solutions. With the LLMs in place, the technology can quickly adapt to new types of calls due to a new product offering or technology changes without new programming which makes GenAI technology a good investment.
While GenAI can help organizations realize efficiencies, it is also important to be sure the models do not introduce bias, especially in key areas like loan approvals. Loans originated with the assistance of GenAI systems must follow the same regulatory framework as any loan. To ensure regulatory requirements are adhered to, it is important that humans, with real-world experience, focus on this, as well as ethical considerations.
Lessons learned
GenAI messaging is pervasive, and the industry is being bombarded with GenAI messaging from all sides. The key to tackling GenAI is to focus on long-term AI strategies and partnerships rather than chasing every new point solution.
A consideration for companies determining whether to pursue AI is having the right resources for development, since GenAI is a specialized skill. Companies may choose to outsource to a company who has a specialty in AI to provide that support. Engage GenAI experts to review the areas that are your pain points and pick one to start with. Even for larger companies, this may be the best way to start.
To identify use cases to pursue AI, companies should review areas where teams of people are doing repetitive tasks. Positions with lower cost structures may be a place to look. Currently the larger companies are more likely to pursue AI, but all mortgage lenders may find that they must adopt AI technologies to remain competitive.
One approach to start with AI is to develop a proof of concept to try a simple scenario and evolve to other areas, exploring where the most value can be realized. For example, extracting data from select documents with a large language model by focusing on one area and expanding as experience grows. Lessons learned from the proof of concept are applied to adjust the process before incorporating AI more broadly.
As technology continues to evolve, the potential for innovation is vast, offering opportunities to streamline operations. As with any powerful technology, it is crucial to approach adoption with careful consideration of ethical, regulatory and security implications and to maintain balance between technology capabilities and human expertise. The future of the mortgage industry, powered by GenAI, promises to be one where technology and human expertise work together to create a more efficient, transparent and borrower-centric ecosystem.
