AI shouldn’t be your only Excel training

Microsoft Excel in use on a laptop.
(Image credit: / Kaspars Grinvalds)

When OpenAI first enabled public access to ChatGPT, not even CEO Sam Altman anticipated this level of chatter. And with the recent launch of ChatGPT Plus and Microsoft Copilot, the noise only seems to be growing. Whether excitement or concern, the impact, both present and future, of generative AI is undeniable.

As these models become more sophisticated and integrated into services we rely on every day, the question remains, is it helpful for day-to-day work right now? The answer for Excel, one of the most popular business apps, is “sort of.”

One of the beauties of Excel is that it is so incredibly simple to use and versatile for different applications. Despite being almost 40 years old, it is still a backbone technology for hundreds of industries and can manage every task from keeping lists to developing demand forecasting supply chain management systems. Estimates say that there are 150 million global business users of Excel and 750 million global users in total. 

The barrier to entry into Excel is incredibly low, but the education process of using Excel can be overwhelming. It has hundreds of functions for layout, data manipulation, arithmetic, trigonometry, statistics, and more. Plus, its applications continue to grow. Without counting special add-ins, Excel has over 500 formulas and even more function points.

Rob Tyrie, head of Partnerships and Customer Engagements, Coherent.

Rob Tyrie, head of Partnerships and Customer Engagements, Coherent.

Most people, unless they learned Excel in finance classes or with specialized training, figure out how to use Excel on their own. They begin with simple tasks and then graduate to items with more complexity. But that process of learning by doing still requires some form of education and lots of it. 

 In my experience, here’s how people learn how to do a new task in Excel: 

  • Phone a friend, or ping a knowledgeable colleague on Slack with your question and hope said person knows the answer.  
  • Follow along the type-ahead hints that help you fill in the parameters you need to calculate something like a mortgage payment
  • Use Microsoft Help (which despite its name, isn’t typically the most helpful tool if you’re searching for a very specific answer)
  • Read the Excel Manual, or get a simpler guide like “A Dummies Guide to Excel”
  • Take a course – there are lots of free ones available, but that takes lead time and patience to cram all that in
  • Do a Google search or watch a YouTube tutorial

But none of those tasks are quick and often are not very intuitive. If you’re lucky, they can get you on the right track, but they aren’t the holy grail answer you’re searching for. Google will use the words of your query to search through a massive index of files to use. It’ll provide you with the closest matches to explore, but it still requires a set of time and research. 

ChatGPT and Excel

AI models like ChatGPT are different. What it does is read a whole problem, make sense of it, and then generate what it thinks is the answer to the problem following the rules and grammar of Excel formulas. It explains the steps in clear English to the point where you can cut and paste formulas to try them on an Excel sheet.

The chatbot is probably not always correct, but it’s typically close enough to try. Information is presented in clear and friendly language, so even if it’s not a perfect solution, it doesn’t waste too much time and could help you troubleshoot. 

Formulas on their own are valuable, but you can also ask ChatGPT other prompts, such as:

  • How do I create a 3-sheet financial model with all the common ratios?
  • Where can I find a data source to add city and state to a spreadsheet where I have IP addresses?
  • How can I use a web service to get live stock price history for IBM in 2021?
  • How do I add a Monte Carlo to a 3-sheet model to test scenarios?
  • What are 5 different types of DCF analysis? Give a summary of each, the formulas used, and an example of usage.
  • What are the pros and cons of VAR analysis?
  • What are good ways of presenting charts of data for stock trend analysis?
  • Make me a 20-question quiz to practice xlookup with financial examples. After the quiz, give the answers with formulas and logical steps.

When I tried ChatGPT, it gave me answers quickly, but I would have learned more if I studied the work and practiced exercises. For education purposes, I know I will strongly recommend my team use ChatGPT to support their work, even if education should be done elsewhere (like in the avenues I mentioned above). 

There are also still questions of accuracy. Despite GPT-4’s ability to pass major exams, the model still has the challenge where it hallucinates. Eventually, these large language models may get to the point where they are actively trained with real business scenarios and can be a game-changer for enterprise software. After all, that’s the beauty of AI – it learns from past inputs to continuously refine and become more valuable. We’re not there yet, but it’s likely not that far off. 

Little is known about how Microsoft Copilot works yet besides a few demos, but given that it relies on the same model as ChatGPT, I would imagine it has to be similar. It will not only make Excel easier to use, it’ll also reduce error rates (a fairly common phenomenon in spreadsheets today). While we can hope for a future where we can all rely on tools like Copilot, for now, we can use ChatGPT to find the answers without too much searching. Just make sure to check your work afterward.

We've rated the best Excel online courses.

Rob Tyrie, head of Partnerships and Customer Engagements, Coherent.