How technology is finally leveling up risk management at banks
From fraud detection to regulatory compliance and credit analysis, tech tools are poised to make risk management easier, faster, and more reliable.
By
One could argue the entire underwriting process is about evaluating risk. Yet, manual underwriting processes make it expensive to evaluate a borrower and certain types of analyses are unfeasible to perform manually, no matter how many resources you have.
For example, 86% of banks say transaction analysis is more important than credit scores for driving small and medium business (SMB) lending decisions, with 9 out of 10 banks employing some form of transaction analysis for decisioning.
However, the majority are doing only a cursory review of activity levels and minimum balance thresholds rather than performing highly detailed analyses because inefficiencies and manual processes within the loan origination system (LOS) prohibit deeper transaction analysis. On the other hand, of those banks that rate their LOS as efficient with SMB loans, 68% are performing highly detailed transaction analysis to qualify and underwrite borrowers.
Technology presents new opportunities for banks to efficiently mitigate risk across three crucial areas: credit risk, financial crimes, and regulatory compliance.
1. Credit risk
Current processes for credit risk analysis in loan origination rely heavily on personal and business credit scores. This leaves many small businesses—namely those that may have a profitable business but poor personal credit or that deal primarily in cash—at a disadvantage. This also leaves banks reliant on the credit bureau’s evaluation of the borrower. To fix this shortcoming, banks should use technology to independently evaluate each variable within the score.
This starts by combining existing account data with data from third-party banks through aggregators like Finicity, Plaid, or Yodlee. These statements can be ingested and turned into machine-readable text that can be used for detailed transaction analysis. Add to that automated KYB checks, identity verification, and tax document pulls, and in minutes, the bank has a deeper analysis of the borrower’s credit risk than it ever had before with hours of manual processes.
AI and machine learning can also be used to improve credit risk monitoring. For example, AI can monitor a borrower’s spending habits, credit score, and payment history and send an alert if those metrics start to indicate potential financial difficulty and a greater risk of default.
2. Financial crimes
Despite multiple due diligence checks in place, fraud still remains an issue for SMB lenders. SMB lenders reported a 14.5% increase in fraud from 2021 to 2022. Sixty-eight percent of that fraud was caught after the point of origination. While a lot of fraud goes unreported, investigations into the PPP loan program found that at least 17% of all COVID-19 Economic Injury Disaster Loans and PPP funds were disbursed to potentially fraudulent actors, amounting to over $200 billion.
While banks have responded to fraud by hiring more staff and tightening restrictions on online transactions, ultimately the bank has to strike the right balance between asking enough questions to ensure the applicant is legitimate while not creating too much friction for the borrower.
The good news is that a robust underwriting solution can layer automated checks of business ID, bankruptcies, judgments, UCC filings, criminal background, and current liens into the customer journey. This, in turn, allows banks to evaluate more fraud risks without creating an experience that’s overwhelming or difficult for the customer to use. In fact, of those banks that use layered fraud mitigation solutions, 50% are able to catch fraud at the point of origination versus the 18% that don’t use a layered approach.
Incorporating transaction analysis into underwriting criteria can also help reduce fraud. If your origination system sees a healthy variation in transactions and no suspicious transactions or statement tampering, then the borrower is unlikely to be fraudulent.
3. Regulatory compliance
Of all their top risk category priorities, 53% of community banks placed compliance in the top three. Staying on top of new compliance regulations is a monumental task. This year alone, changes have been made to SBA and Community Reinvestment Act rules, and CFPB 1071 added new data collection and reporting requirements for SMB lenders.
Technology with natural language processing capabilities can help banks stay on top of changing regulations by analyzing documents and extracting key points of information.
Automated underwriting systems can help banks collect legally required information under CFPB 1071, such as gender or veteran status, by collecting and storing the data at the time and place that makes the most sense for the customer.
Within compliance, 79% identified fair lending and 35% identified third-party or fintech partners as top priorities.
While disparate treatment can be fairly simple to avoid within risk policies, disparate impact can be harder to account for. For example, credit scores are commonly an ineffective tool for measuring the creditworthiness of underserved communities. Studies support creating a well-rounded and well-documented approach for evaluating borrowers, which can help reduce the potential for bias and create fairer outcomes for those communities.
Bank risk management as a revenue-generating function
While automation can improve efficiency and reduce costs across the three areas discussed, perhaps one of the most interesting use cases for technology in bank risk management is revenue generation.
Many current risk models are binary—if a borrower barely misses the minimum monthly revenue threshold, it’s an automatic rejection, even if other factors indicate a good risk potential.
For example, take two SMB borrowers with the following profiles.
| Borrower A | Borrower B | |
| SBSS Score | 175 | 165 |
| Deposits | 4-6/month | 10/month |
| Minimum Balance | $5,000 | $10,000 |
| Annual Revenue | $250k | $300k |
Historically, banks might have a fixed SBSS threshold of say 170, meaning that Borrower A in this example would receive funding, while B would not. But in fact, the data would say that Borrower B is actually a very low credit risk.
Now, imagine a more complex profile developed by regression analysis of past loan approvals optimized for expected outcomes. By using a more holistic approach, the bank is able to adjust its minimum thresholds and optimize how it evaluates risk, all while potentially approving more borrowers with positive outcomes that may have been overlooked in simpler models.
In this scenario, risk management technology transforms from a pure cost function to a tool that can be used to drive new revenue for the bank.
Low risk, high reward
It’s a rare opportunity for a financial institution to simultaneously reduce risk, improve the customer experience, and drive new sources of revenue. However, as banks evaluate priorities and budgets, reframing how they think about risk management could have a huge impact on the bottom line.
Learn more about automation in lending.
