Is AI taking over the world? Listening to some people, you might think so.
But – slow down! Not so fast.
We were fortunate to attend a panel discussion on “AI and the Banking Industry” at the recent NAGGL conference. As the panelists explained, there are some areas where it makes complete sense to take advantage of the efficiencies AI can offer. And then there are tasks and activities where the human touch remains essential.
Let’s cover a little groundwork, and then review few examples.
The first crucial step is to make sure you have one central owner for any AI use. Letting anyone who wants to give AI a whirl is begging for problems. And there should be someone on the team who understands how to craft AI queries – it’s not as simple as you might think – and SBA loan packaging.
Secondly, start with small pilot project. Pick something simple and straightforward to test out not only the technology, but also how your team will interact with it. And make sure your team gets training!
Potential areas where AI can play a role:
Borrower eligibility checks
Compliance verification for SBA requirements
Warnings about potential defaults, based on predictive analysis
Fraud detection
This is, of course, a very high-level list that only barely touches upon the many possibilities. And any use of AI must be considered relative to risk factors if something goes awry. The pilot project should include multiple levels of human verification – as should any additional projects going forward.
Areas where AI should not play a role:
High-risk decision making – this sounds obvious, but make sure the AI processes don’t wander into areas where they’re just not qualified.
Interpretation of SBA policy changes or requirements updates – the AI tool can’t interpret the nuances, and you don’t want to end up out of compliance.
Final decisions on credit – AI might be helpful in initial scoring, but it should never make final decisions.
Customer interaction and relationship-building – this is one where there are multiple opinions, but in our opinion, there are too many ways customer interaction can go wrong. You know what we mean, because you’ve experienced it!
Fraud detection without human intervention – AI can certainly identify anomalies that might point to fraud, but there are legitimate exceptions that only a human can recognize.
Clearly, this is – again – a very high-level list. And we can’t possibly identify all the risk factors in using AI tools, any more than we can identify all the possible gains for using AI tools.
In short, we recommend:
With the caveat that we are not in any way AI experts (or even AI novices!), we can clearly see that there are some significant benefits to smart use of AI in SBA lending. And let’s face it, if you don’t take advantage of AI, your SBA department is likely to start falling behind the competition.
So start with that pilot project.
Make sure you have smart, risk-averse people leading the way.
And as the old saying goes, trust but verify: trust that your AI experts, and therefore your AI tools, are doing the right thing, but verify the results – and not only in the testing stages, but on a regular basis; it should be part of your checklists and review process. AI models change, SBA compliance requirements change, and AI is well known to have biases – and to “hallucinate” with incorrect answers.
While we’re not qualified to offer any further advice on AI specifically, we can provide reviews for loans where AI was used. We’re happy to be part of the pilot project verification process, as well as ongoing validation that all of SBA’s requirements are being met. As SBA consultants and SBA trainers, we can give you the assurance you need that your AI process is working as desired.
Give us a call at 877-576-0819, or drop us a line through our Contact form here!
Comments