Artificial Intelligence
Dec 12, 2018

Why the value of AI-assisted loan underwriting should no longer be unsung

The merits that artificial intelligence (AI) can bring to the loan-underwriting process deserve greater appreciation. AI-assisted loan underwriting draws together big data, which constitutes 20% of the information available to underwriters, as well as mixing in social, business, and internet data. It can also fold completely unstructured data into the mix so that decision-making is a lot easier and much more effective.

AI-assisted loan underwriting is ultimately about bringing in additional data sources and technologies to improve the overall quality and speed of the loan-underwriting process. It's about blending together traditional and alternate data sources, as well as foundational and transformation technologies.

This type of loan underwriting combines traditional sources, such as Dun & Bradstreet, Equifax, and Graydon, and alternate ones, such as news media, blogs, and social data from the likes of Facebook and LinkedIn. This gives underwriters a 360-degree view of the applicant. AI-assisted loan underwriting also merges foundational technologies like SAP and People Soft with transformational technology like blockchain, AI-enabled services, and robotic process automation (RPA).

Now, decision-making no longer needs to depend on what the underwriter thinks in the middle of a bad, judgment-clouding day. On the contrary, misperceptions are corrected over time based on what all the new, AI-supplied data is telling the underwriter. As internal big data is just 20% of the available total, adding new inflows has a tremendous, transformational impact. By bringing the 80% of unstructured data into the equation, underwriting decisions are made faster, simpler, and better. Much of the 20% is data that's disparate, non-standardized, and difficult to access. But properly designed AI solutions can bring it into focus, too.

Asking the right questions

To facilitate the AI-assisted loan underwriting process, it's vital to ask the right questions. One of the biggest questions is, "How should the legacy process combine with the digital?" But along with this question, it's equally important to ask, "Why am I doing this? Do we really need to run this process, or can it be augmented by another step in the life cycle that's AI-enabled?"

In asking these questions, it's useful to remember that AI allows you to be proactive rather than reactive. AI also reduces redundancy by eliminating unwanted steps. For example, instead of waiting for year-end or quarterly financials, why not ask the businesses for their bank statements? Even better, ask them to provide access to their credentials (similar to what wealth managers do). Bank data offers a sense of what businesses spend, and analyzing this data can provide information about their health. So, the process becomes more real time rather than after the fact, and more proactive.

Organizations attempting to build an AI-assisted loan-underwriting system must also consider that sometimes legacy systems are so arcane that the prevailing attitude is, “If it ain't broke, don't fix it." But the reality is that if you don't change, then your fast-evolving competition will overtake you, which is why it's crucial to bring AI in alongside foundational technologies.

Keeping the customer as the focus

In moving toward an AI-assisted process, think of AI as a wheel with a customer experience (CX) focus at the center and spokes of AI, automation, and analytical capabilities. In bringing all available technologies together – machine learning, deep learning, computer vision, conversational AI, RPA, and more – remember to design with the end goal in sight of the right CX core.

Keep in mind that if loan underwriters ask for six documents for approval, they still need technologies that can go back to learn what was asked of the customer. Some use cognitive computing to do this, others natural language processing, and still others computer vision. But it's when you bring the full range of available technologies together that you have a very powerful, augmented AI tool.

The CX goal at the center brings customers to the table and gives them what they're asking for before they ask. Conversational AI, for example, provides spend analysis of how the customer transacts with counter-parties. Or, for smaller businesses, you may be able to rely on bank statements and tax returns. The bottom line is, if your AI solution helps speed the data review process, then you've invested wisely.

It's also beneficial to put RPA in the mix to work around legacy technology. Together, all these digital technologies help realize an end-to-end representation of the data. What best determines whether your AI solution is built to last is what the CX looks like. Are loan underwriters and customers both getting the experience they want and need? That's the big question. The answer should be yes.

To learn more about this topic, you can listen to a recent webinar called, "Transform your lending firm into a digital lender of the future."

About the author

Srini Bharadwaj

Srini Bharadwaj

CTO, AI products

Follow Srini Bharadwaj on LinkedIn