Will ChatGPT Be My Banker?

ChatGPT (and AI programs like it) are powerful tools that have the ability to change the way bankers do business

Over 1.8 million professionals use CFI to learn accounting, financial analysis, modeling and more. Start with a free account to explore 20+ always-free courses and hundreds of finance templates and cheat sheets.

AI in Banking: A Historical Perspective

It was not so long ago that bankers transported physical cash and wrote loan and account balances in paper ledgers. It was an era where those bankers were often the only reputable citizens in many towns to talk to about finance-related topics.

Since that time, advances in technology have changed things (not just in banking, of course). But no matter what distribution channel a client uses to engage with financial services today (ie. branch, telephone, online, etc.), there are still many human elements. At the very least, clients know that a human can be reached if a technological glitch were to necessitate it.

But is it possible we’ve arrived at a juncture in history where this may no longer be the case?

Disruption Today

Enter AI (artificial intelligence), and we have a new era of professional disruption upon us. 

What’s puzzling to me is how openly people seem to be embracing it (even if just out of curiosity). Truth be told, I was frightened when I heard about ChatGPT’s capabilities and the rate at which it seemed to be achieving total ubiquity.

It seems obvious to me that a best case scenario is ChatGPT (and other AI technology) changes the way that tens (or maybe hundreds) of millions of white collar workers do their jobs; the worst case is that many of them no longer even have jobs in the relatively near term.

Why It’s Personal

I’m no longer in banking, but if the bankers we train through CFI’s enormous content library end up out of work, that presents an existential professional threat for me, too. 

Ditto if ChatGPT or its counterparts can replicate some meaningful amount of our IP, or learn to teach more effectively than my colleagues and I (neither of which is all that far fetched). 

As something of a “traditionalist,” the idea of being complicit in my own professional demise seemed dumb and ill-advised. So, you can understand my apprehension around digging into this technology and helping it learn faster.

But after fighting it for a while, I finally caved in and decided to see for myself what all the hype and saber rattling is about. And the irony was not lost on me that, when I created my OpenAI account, Cloudflare made me confirm that I’m human… really?!

Anyhow, I decided to tap into the powers of ChatGPT to answer one simple question: 

Is ChatGPT more likely to help bankers or render them obsolete? (And by extension, myself!)

AI in Commercial Banking: Case Study #1

My background is in lending and that’s what the CBCA program is all about, so it seemed a reasonable place to start. And since AI is renowned for combing through reams of information really quickly, I wanted to tackle financial analysis first.

We have financial statements for dozens of example companies that we use to teach different ratios and other topics, so I fed ChatGPT just the following balance sheet from one of those example companies and asked it to weigh in on the financial health of the business.

balance sheet

Here’s what I got back:

ChatGPT analysis of a balance sheet

Here’s my initial take:

ChatGPT provided the fastest elevator analysis I’ve ever seen

But at the end of the day, it was merely that — elevator analysis. This is when we go through the financial statements and say: X is up year-over-year, Y is down, and so on. The analysis lacked substance.

I suppose we could use that information to inform the nature of questions we might want to ask a management team when we sit down with them, but that’s hardly reinventing the role of a business lender. 

But wow, it was fast!

ChatGPT made surprisingly subjective observations

For example:

  • “The company has a significant amount of trade and other receivables…” — I would ask, “significant” relative to what? There was no benchmark information included for comparison.
  • “The company has a substantial amount of long-term debt…” — again, “substantial” relative to what? 

The commentary reminded me of the kind of submissions junior analysts put together in their first half dozen credit applications, before they start to understand what they’re doing or what they’re supposed to actually be looking for.

ChatGPT arrived at some odd conclusions

For example:

  • “Total assets have increased…indicating an improvement in the company’s financial position.” Personally, I don’t agree. Growth in assets, absent any other information (like relative to sales, at minimum) doesn’t tell us much. Further, this company had a zero cash balance at the end of both years; this would have been flagged for further investigation by a human with any lending experience.
  • “…more information is needed to determine the company’s profitability and cash flow.” That’s technically correct, but lacks nuance. For example, you could make a pretty reasonable estimate of profit by subtracting the previous year’s retained earnings from the current year’s; of course, that doesn’t account for dividends, but I like to think that a human with credit experience and decent financial acumen would simply state that caveat.

But I wanted to give it the benefit of the doubt…

So I gave it the information needed to derive a cash flow statement using the indirect method and asked it to do just that. I stipulated that there were no dividends and clarified all non-cash items, including depreciation expense. 

And you know what? It came pretty close, off by only $2,228. But there were several mistakes that added up to $2,228 and, in the time it took me to figure out what these were, I could’ve built the cash flow statement correctly myself several times over.

It was impressive to see how close it came. But probably the most important point is that it doesn’t really matter since I already had a cash flow statement and this was just a thought exercise. 

Financial analysis is only partly about making quick calculations; the crux of financial analysis is really about deriving useful insights and the AI fell well short of what I’d consider nuanced or actionable insight.

AI in Commercial Banking: Case Study #2

Given an apparent inability to generate quality insight, not to mention what I expect to be unprecedented regulatory overhangs, it seems unlikely that ChatGPT will be replacing human interventions in commercial credit (and by extension anyone on the relationship teams at traditional commercial lending operations) any time soon.

But I wanted to explore if there are ways it could support some of the qualitative work we do. As an industry, we do a lot of business development and relationship management, after all.

So I put my banker hat on and pretended to be prospecting the very company I work for — CFI. I started by asking what sort of questions I might ask a business owner in the eLearning industry. Here’s what I got:

ChatGPT analysis of the eLearning industry

My thoughts:

These are actually very good question areas

If I wanted to impress an owner or a senior leader in the eLearning industry, these topic areas would be a terrific starting point; all are absolutely top of mind. 

The idea of teasing out challenges and opportunities isn’t new or revolutionary, but the fact that it identified content creation, learning management systems, and target market/pricing questions is fairly astute.

I don’t think it’s better than existing platforms

Many industry research platforms currently exist that support sales professionals with industry-specific call-prep resources (for example IBISWorld and Vertical IQ are both popular in the banking community). 

Is ChatGPT providing more nuanced insight than those? My couple of hours playing around with it was an admittedly small sample size, but I don’t think so. On the other hand, the fact that it’s dynamic and you can drill down into specific sub-topics of interest is a very compelling value proposition.

As a sales/relationship management professional in banking, you could learn an awful lot in a very short period of time, about any industry or topic you want.

And, ChatGPT is free!

Even though the technology isn’t (necessarily) producing highly tailored, topic-specific call prep questions in line with incumbent platforms, it is free. 

For small firms or for independent loan brokers or advisors that don’t want to pay for research platforms, ChatGPT could be a very useful starting point for industry information and call prep resources.

But I wanted to dig a little deeper…

So I started asking questions about CFI, specifically, and some of the important individuals involved in the company.

I’ll spare our readers screenshots of all the results, but suffice it to say this was a mixed bag. It had some facts correct, but there were also some inconsistencies including timelines, key players, “so-called” partnerships, and others. 

I suspect this tool may be much more accurate in digesting and synthesizing public company information, since vetted and accurate 10Ks and MD&As are available for online consumption. When it comes to private, smaller, and owner-operated businesses, AI-generated results should be taken with a grain of salt since, in general, there’s very little material information available in the public sphere. And these types of business borrowers make up an enormous proportion of many loan books.

So?

It’s no surprise that a number of America’s most notable firms have already barred employees from using ChatGPT in the workplace, including CitiGroup, Bank of America, and Goldman Sachs; citing accuracy issues and the potential of disclosing confidential information, among others[1]

The bar in terms of what’s fact and fiction on the internet is already pretty low; it’s important that critical thinking be applied and the accuracy of AI-generated results remains a question, at least for the time being.

Will it get smarter? Yes — which is itself both exciting and terrifying. But we aren’t there yet and it’s likely to take a long time before financial institutions are willing to stake their reputations on generative AI.

AI in Banking: Fact or Fiction

I wanted to close out with a fact or fiction exercise to help hammer home a few key takeaways.

AI is new to banks (and financial services firms more broadly)

This is fiction

Machine learning (ML) is a form of artificial intelligence and it’s been a cornerstone in most risk models for years. 

Machine learning is a process that uses statistics to help spot patterns in data that can be used to make predictions. CFI even has a course on using machine learning for loan default prediction that’s several years old.

Other AI use cases are also used in financial services, including product recommendations through online banking for retail clients (based on previous spending behavior) and automatically generated limit increases (based on a client’s historical repayment patterns). 

These were on the scene long before ChatGPT!

AI is capable of performing complex tasks that once required input from humans

This is a fact

This is what makes AI so incredible, but also where a lot of peoples’ fear emerges from. But technology has been automating previously manual human tasks for a long time; even in financial services.

We need not look any further than Microsoft Excel to see a prime example of complex tasks becoming vastly simplified — and the folks using Excel now spend more time analyzing information and deriving actionable insights from their models, rather than working through rote and tedious calculations.

I don’t know any finance or banking professionals that would trade in their spreadsheet software, do you?

AI is going to make bankers obsolete

This is fiction

Like the Excel example, along with countless other examples like Bloomberg terminals, APIs for easy system integration, and even email, technology (including AI) is potentially going to change the work that we do; it seems less likely to me that it will result in massive front line layoffs.

Why? Because several truths exist in banking; most of which have remained incontrovertible during my time in the industry. These are:

  • Client love = leverage. A great piece of advice I received early in my career was if your clients love you, you have leverage

While the banking “relationship” is technically owned by the financial institution (by way of “sticky” credit and cash management products), you can absolutely take ownership of how you manage your relationships with the humans that operate those borrowing companies.

And if your clients love you, you’ll have negotiating leverage across a variety of dimensions because some key clients could become a flight risk if you move on to a new FI. 

  • Regulation rules the day. Especially for chartered banks, which is why many are banning the use of generative AI by their staff. 

Could this technology possibly support regulators in scanning audit files more quickly to flag potential issues? For sure.

Could this technology fit nicely into fintech platforms to support an initial screening for, say, small business loans or consumer credit? Absolutely (in fact similar technology already underpins the infrastructure of many fintechs and it has for years, it’s just better now).

But I’m confident that this technology won’t be imminently replacing loan officers or adjudicators, since protecting depositor funds and the health of the broader financial system is far too important to turn over the reins to machine learning technologies.

  • Connection is critical. While human beings love speed and efficiency, when it comes to the financial health of their business, it’s still really nice to be able to pick up the phone and speak to another human.

AI can’t have the kind of compassionate conversation that’s required when management breaches a covenant or falls on hard times and has to discuss restructuring credit. AI can’t take its client out for lunch or a round of golf to celebrate a funding or some other business success. 

The best way to future-proof a career is to both connect people and connect with people at every opportunity. Technology has paved the way towards much more of a service-based economy than we had pre-industrial revolution, and serving humans effectively will serve you well in the still-evolving service economy of tomorrow.

Additional Resources

How Finance and Banking Professionals Can Use ChatGPT

Big Data in Finance

7 Relationship Lessons I Wish I Learned Earlier in my Banking Career

See all commercial lending resources

Article Sources

  1. Workers’ ChatGPT Use Restricted At More Banks
0 search results for ‘