3 Reasons Why ChatGPT and Other AI Tools Cannot Replace (Good) Data Analysts

???????I remember the 1st time I experimented with ChatGPT; I was at a Christmas party in December 2022 when one of my good friends who owns a tech company, told me about this awesome tool that would usurp Google as the world’s best search engine. Encouraged to try it out, I took off my tuxedo jacket and immediately created an account on my phone.?To say I was blown away with the results would have been an understatement, and it appears I wasn’t alone in the admiration. 2023 has been the Year of AI with discussions about how it is going to change the world. One of the chief ways AI is predicted to change the world is through the elimination of jobs due to efficiencies gained. The list of jobs AI is projected to impact run the gamut from lawyers to doctors to marketers. On a more personal note, AI is projected to deeply affect my field of work which is data analytics. As someone who has used ChatGPT heavily, these are 3 reasons why I don’t foresee ChatGPT and other AI tools replacing (good) data analysts.

1.??????One Needs to Competently Know the Data Analysis Task to Use ChatGPT Most Effectively: ChatGPT and other AI tools present an interesting quandary in that, to use them most effectively for data analytics support, you need to have a good idea of how to do the task at hand. A novice or even a lowly skilled data analyst could not obtain good results from ChatGPT, as the quality of your prompts in ChatGPT are directly corelated with the depth of your understanding of the task requirement. For example, as part of my data analysis / visualization work with my clients, I must write a summary of the tasks that were performed as part of a monthly report. I had just created a process flow diagram, to help my clients understand the various steps that must be followed to create an effective PowerBI analysis dashboard. Rather than type up a report that would have taken me about 2+ hours (more if I had assigned it to a junior resource), I decided to employ the use of ChatGPT. I opened ChatGPT and wrote the following “Create a process flow for how one would go about creating a procurement dashboard in PowerBI, with emphasis on the stability of the data source.” ChatGPT gave me 7 factually accurate, but highly generic process steps.?I then spent 15 minutes editing and customizing the results to be more specific to my client’s environment, and I had saved myself several hours drafting a report. My report was successful because I already knew the result that I wanted and knew how to create it if needed. What I was seeking to do was to become more efficient, and ChatGPT was very helpful in providing me with an 80% solution. A novice would not have been able to accomplish this task, as they would likely not know the right questions to ask. Additionally, it was my deep knowledge of my client’s environment that helped me customize the responses to be something meaningful to my client. It was my familiarity of PowerBI and my client’s environment, that made my deliverable meaningful.

2.??????ChatGPT Has Not Yet Mastered Nuisances in Data: In my spare time, I run a TikTok page titled @DCDataGuy, where I primarily analyze sports phenomenon using data. As track athlete, I wanted to explore the claim that 100m sprinters are getting taller and heavier over the years. Rather than spending hours looking up data in Google, I asked ChatGPT to provide me with the name, citizenship, height, and weight of every Olympic 100m gold medalist since 1920. Predictably, ChatGPT provided me with the requested data, providing me with the names of these gold medalists on a four-year basis, but the results stopped in 2016. I ran the tests at least 3 more times but kept getting the same results. I knew the answer; it was Marcell Jacobs from Italy; however, the 2020 Olympics were conducted in 2021 due to the 2020 Covid shutdowns. Without knowing the engineering behind ChatGPT, my guess is that it detected a pattern of the Olympics occurring every 4 years but did not understand why the 2020 Olympics were conducted in 2021, so rather than providing me with a result, it just ended at the last regularly scheduled Olympics in 2016. While ChatGPT is a highly effective data gathering tool, this instance made me question its ability to understand very specific nuances in data. There are times when business processes must occur outside of a regularly scheduled period, and it begs the question of whether many of these AI tools have the capacity to understand and measure these very fine nuisances, so much so that they could take the place of data professionals.

3.??????AI Could Rob People of The Joy of Competently Building Things: As part of my job, I help my clients with gathering their user requirements and helping them make sense of their data. Many of my clients don’t often know exactly what they want, rather, they have a hypothesis or a directive from a superior, that they need to bring to life. I have witnessed on several occasions when a confused customer comes up to me with a request, and then after some exploratory sessions of structuring, cleaning, analyzing, visualizing, and reporting on their data, said customer has gained a deep sense of satisfaction and knowledge from undertaking the process. Even if AI tools were able to provide users with the exact data analysis that they needed, humans are innately attracted to building things. We have a natural curiosity and creativity, and a deep desire to solve problems. Many of my clients could simply provide me with their user requirements and be done with the ask, however, they often have a deep desire to work with me in the building of their projects, even though they are not the ones doing the building. This is because building is a means of solving problems. Whether it is a shelter to protect us from the elements, a bridge to cross a river, or a tool to help us work more efficiently, building allows us to find solutions to the challenges we face.?AI tools could help us with the creation of the data product, but the joy lies in the building.

For full disclosure, 3% of this essay was obtained from ChatGPT.