Nearly any upper-level financial leader who was required to take a 101-level Computer Science class in college has heard the phrase “garbage in, garbage out.”
In essence, poor or erroneous raw information or data is only going to lead to a sub-standard end product. It’s akin to cooking a meal with a bunch of expired ingredients.
This long-standing concept weighs heavily when discussing the potentially sizeable increase use of generative AI (GenAI) in finance functions. One of the key concerns we at Emagia have heard from customers and colleagues at the CFO level and at industry events like NACM Connect’s recent Gateway Conference in St. Louis for credit managers revolves around the potential for erroneous information to contaminate any GPT solution. As we know, there’s a lot of garbage – in the form of misinformation – out there — “just because it’s on the Internet doesn’t mean it’s true.”
So, to participate in the wave of GenAI in finance, does that mean simply surrendering to the mercy of the Internet and hoping it just works out?
It doesn’t have to be this way… no should it be.
Trust Yourself As the Source
With the massive spike in notoriety over the last year of Open AI’s Chat GPT solution, the development of similar solutions has skyrocketed. And though early GPT solutions pulled data from all corners of the Internet, and increasing number of more discerning options now exist.
Using a closed-cloud approach – like Emagia’s GiaGPT tool for autonomous finance – drastically reduces risk because the ultimate user knows the information being used is credible. That’s because with a closed cloud approach, the financial professional (or other user) simply uploads the documents or spreadsheets they want to pull information from or use as a source from their own company’s files as the source material.
As such, misinformation for less reputable Internet sources does not infect your research because you’ve already vetted it.
Of course, not all AI-based solutions are created equal – or developed by teams with years instead of a few months of actual experience in autonomous finance. So, vetting the solution and the provider to eliminate the fly-by-night developers remains critical.
So Should CFOs and Other Financial Leaders Avoid ChatGPT Entirely?
Simply put: no. ChatGPT broke down the doors for AI adoption, especially in the United States, and it’s far from on its last legs after less than one year in the public sphere.
The key is all in how you use it. Basically, it’s all in the prompts.
Remember, we started with the “garbage in, garbage out” anecdote. The applies to the user of ChatGPT as well.
Taking the time to carefully craft your prompts (re: what you’re asking ChatGPT to do for you) helps to maximize results and minimize potential problems can prove incredibly valuable.
For example, if you want to ensure the data are trustworthy, ask ChatGPT to pull information from only from certain, trusted sources – be specific. Or go so far as to literally copy and paste bullet points, stats, an executive summary or entire reports you want to source and ask ChatGPT to pull only from those as part of your prompt.
It may seem like overkill, but it works in mitigating the risk of false information and data infecting your ability to confidently make fast and bold decisions.
There’s no reason to give up on ChatGPT just because the Internet is fraught with lies and misinformation. Just remember, especially in your early days of adoption, to put a little extra thought into what your true goal is and what you want ChatGPT to do. And, of course, double-check the work for things that look amiss.