Spreadsheet software concept in green and blue colors illustration, can be used in animation, … (+)
As part of an ongoing series, I’m trying to put theory into practice and help everyone who reads this column apply AI to their daily professional lives. In my first article I covered how to get started with text-to-speech technology in a few simple steps. Today I’m going to cover data analysis, which is going to be an increasingly popular use case (which is not without its flaws).
How does AI work with spreadsheets?
You’re probably used to having a data analyst on your team who is an absolute spreadsheet expert and who is your go-to guy when you need to do custom Excel work. I used to be that person – using a combination of Excel, MS Access (with SQL), and tools like SAS to do big data analysis. But now with AI, the process is much simpler: your best analyst can have much greater leverage, because they won’t need to waste their time on unnecessary tasks like data cleaning .
Here’s a general idea of how AI works:
- Data analysis: AI deciphers the structure of your spreadsheet, identifying rows, columns, headers and data types.
- Pattern recognition: Using machine learning, AI recognizes patterns and correlations within data, even those that are not immediately apparent to human analysts.
- Data cleaning: It automatically detects and corrects errors or inconsistencies, such as missing values or outliers.
- Insight generation: By applying statistical models, AI obtains insights, forecasts trends, and makes predictions based on the data.
All of this is to say that a great analyst who knows how to ask the right questions will generate more insights than an analyst without AI.
How can I try this?
- Create a Chat GPT Plus account. There are other options available to you here, but since most of you are probably familiar with Chat GPT, go with what’s familiar.
- Find the “Explore GPTs” tab in the left pane. Once you’re there, select the GPT Data Analyst – pictured – and you’ll open a chat interface with him.
- Select the dataset you want to work with, download it in CSV (or any other format). If you’re looking for sample data to work with, head over to Kaggle for a great repository with lots of cool stuff.
- Upload the CSV to the chat window and invite Chat GPT to interact with the spreadsheet. I often use the prompt: “Please read the CSV I am about to upload, clean it of any errors, and prepare for a series of questions I will ask you for in-depth analysis.”
- To ask questions. This is the key: ask the right questions. Examples might include “Discover 5 non-obvious insights from this data.” “Help me spot long-term trends that I might be missing in some way.”
In the example below – I downloaded a dataset on New York real estate from Kaggle. I then uploaded it to Chat GPT and asked him to find out some information. Here’s what he found:
Example of Chat GPT output from a data analysis task.
Sure, correlation analyzes aren’t very deep, but without a lot of extra incentive, they became much more relevant very, very quickly. One obvious gap I’ve noticed, as a disclosure, is that as data sets become larger, they can sometimes struggle in terms of speed and accuracy of analysis.
How should you use data analytics today?
There are a number of practical use cases for data analytics today that you should consider using. Examples include:
- Financial forecast – ingest your existing financial data to see if there are trends such as seasonality that you can identify more easily with AI.
- Better HR practices – reduce employee churn and retention and improve your recruiting processes by identifying what’s going wrong.
- Competitive analysis – monitor the competition by evaluating the SEO strategy of your competitors, etc.
The options are sort of limitless, but it would be a huge failure if you didn’t start playing with some of these tools sooner. They will only get exponentially better from here.
