Accenture Banking Blog

In the mortgage industry, early adopters of document automation technologies were underwhelmed with the results of their investment. Today’s advanced AI solutions are changing that perception. They have little in common with early optical character recognition (OCR), which was able to classify only very simple and structured documents. The latest advancements in AI can cut the lengthy mortgage origination cycle significantly and reduce errors.

Mortgage lenders have until recently not benefited from the AI-driven increase in speed and efficiency that has occurred elsewhere in the financial ecosystem. A mortgage typically facilitates one of the most important purchases of borrowers’ lives, so they don’t expect to close their loan in a matter of minutes or hours, like they do when getting approved for personal credit. However, they do expect efficient service and are frustrated when back-and-forth paperwork delays take weeks. A growing proportion of homebuyers are from a generation of digital natives, who expect transactions to happen without hiccups. Capturing the fickle millennial market requires lenders to compete on experience, quality of service, and speed.

To tackle this problem, lenders need to take advantage of the latest AI technologies—particularly from machine learning and natural language processing vendors that are offering vastly improved OCR. Today’s technology can read and identify information in documents at accuracy and confidence levels never seen before.

What document AI does for mortgage lenders 

After decades of slow development, AI has evolved into a far more powerful tool in the last few years. This is mainly due to the expansion of what is referred to as machine learning neural networks—algorithms designed to mimic how the human brain processes information. As computing power has grown, so has the sophistication of these networks and their ability to automate more complex tasks. 

Most everyone is already familiar with neural networks as they are used in facial recognition, virtual assistants like Alexa and Siri, and self-driving cars. They have many other uses, one of which is to recognize information in mortgage documents, identify and classify the document types, and extract key data elements from the documents. These modern solutions have evolved such that information can be identified even if it is organized in different ways. For example, they can now process bank statements from multiple banks and identify the required data elements with impressive accuracy. 

While AI advancement is the primary driver of the improvements, they are fuelled by a substantial increase in computing power using cloud technology. Fintech providers and major cloud service providers have created pre-built AI models that operate in the cloud, trained on millions of pieces of data. Businesses can use these to harness the power of AI with minimal effort and technical capability, so they can put the technology to work for them faster.  

How document AI can transform the mortgage cycle

The value of document AI to the mortgage industry is simple: it automates document processing. Think about the countless hours spent reading documents to identify the document type and then indexing them into the appropriate files, or reviewing income documents and keying in the data. Document AI can automate indexing, extraction, and filing, sparing lenders the tedious business of manually processing most documents. Instead, staff can review only the documents that are identified as problematic by the AI system. 

This will help mortgage lenders in several ways: 

Faster time to close: Most lenders implementing the latest OCR technologies with AI can expect to see at least a 30% reduction in document processing times. This automation alone should shave several days off the average origination cycle. 

Better use of staff time: Most lenders struggle to keep their senior processors and underwriters working on the highest-value activities for credit decisioning. Paper processes continue to use too much of their time for routine tasks. Loans often involve at least 20 document types totaling several hundred pages. When AI is used to pull routine data out of the documents, specialists have more time to do their core, high-value jobs. 

Fewer human errors: New AI models for document processing are extremely accurate, with error rates typically below 5% for structured documents. This reduces the risk of costly data extraction errors by humans that slow down the underwriting process. During especially busy periods, human error rates increase as the workforce struggles to keep up. A machine-driven process delivers consistent results, regardless of volume. The latest AI engines incorporate feedback loops that reduce error rates even further as the model refines itself.

Scale up and down: An often-overlooked benefit of automation is its ability to seamlessly scale up and scale down as demand changes. When the market is booming, the talent required to scale up quickly becomes difficult to find. Document AI reduces and simplifies manual touchpoints, so that businesses don’t need to increase and then reduce staff numbers significantly as volumes change.

How to optimize document AI for your business 

Many mortgage businesses recognize that they would benefit from adopting up-to-date AI but aren’t sure how to smoothly integrate it into their business processes. Follow these steps to adopt and keep up with the latest document AI. 

  1. Develop a vision and a roadmap for the future. Have a clear understanding of how document AI will be part of your business, including funding, operating model, and technology.  
  2. Evaluate document AI vendor solutions. Assess and identify the right technology solution to fit your use cases. Consider specific technical capabilities, vendor hosting models and pricing structure, and your in-house skill sets.  
  3. Transform your workforce. Review people’s roles in light of the automation of many of their previous routine tasks. The workforce should understand how AI will let them focus on more complex and higher-value loan processing requirements.
  4. Continuously evaluate and adjust your roadmap. AI technologies, regulations and the competitive landscape are rapidly evolving. As new value-driving capabilities emerge, lenders must be willing to pivot and adjust direction. They must have a nimble mindset and recalibrate their roadmap accordingly.

Mortgage lenders that adopt the latest document AI tools will be able to automate mundane, time-consuming data classification and extraction tasks. Increased automation will in turn lead to faster loan cycles, fewer errors, increased efficiency, and lower costs.   

To learn more about how Accenture helps mortgage lenders lay their foundation for the future, visit: 

Accenture Credit Services 

Accenture Song 

With thanks to Matthew Dow and Matthew Orr for their contributions to this article.

Disclaimer: This content is provided for general information purposes and is not intended to be used in place of consultation with our professional advisors. Copyright© 2022 Accenture. All rights reserved. Accenture and its logo are registered trademarks of Accenture.