Important things to know
An honest look at how AI tools are quietly changing the way analysts work and how to actually use them well.
Let me give you a confession here. When I heard someone say that I should just rely on AI for data tasks for the first time, I scoffed in their face. For years, I've learned how to use SQL.
But then one day, I found myself unable to fix an especially ugly nested SQL query. In my frustration, I threw it in front of Claude and demanded he find the problem for me. And he did so faster than it took me 15 minutes to try to debug it myself.
This is when my mind changed.
It's Not About Replacing You. It's About the Boring Stuff.
Here's the thing nobody tells you when you start in data: a huge chunk of your time doesn't go into analysis. It goes into setup. AI tools like ChatGPT and Claude are genuinely excellent at all of that. They handle the repetitive, low-creativity parts of the job so you can focus on what actually requires your brain, things like understanding the business problem, interpreting the results, and making decisions.
I want you to think of it less like having a genius colleague and more like having a very fast, very patient assistant who never gets tired of writing VLOOKUP formulas.
Where AI Actually Helps in a Data Workflow
1. Writing and Debugging Code Faster
This is the obvious one, but it's worth being specific about how useful it actually is.
You can describe what you're trying to do in plain English and get a working code snippet back. Want to calculate a 7-day rolling average grouped by region? Just say that. You don't have to remember whether it's grouping sets, rollup or window functions or dig through Stack Overflow for twenty minutes.
A few things I use this for regularly:
- Complex SQL joins and subqueries
- Reshaping dataframes in Python using groupbys
- Setting up matplotlib or Plotly charts with specific formatting
2. Cleaning and Understanding Messy Data
Messy data is the universal experience of every analyst, everywhere. You get a spreadsheet with merged cells, inconsistent date formats, column names like "Column 7," and values that are sometimes "N/A," sometimes "n/a," and sometimes just blank.
You can describe the mess to an AI and ask for a cleaning script. You can paste a sample of the data and ask it to figure out what the columns probably mean. You can ask it to suggest what validation checks you should run before trusting the data.
It won't clean the data for you 100% but it dramatically speeds up the process of figuring out how to clean it.
3. Writing Formulas You Half-Remember
We've all been there, mixing up syntax. You know there's an Excel or Google Sheets formula that does exactly what you need. You remember it starts with INDEX and involves MATCH somehow, but the syntax is just... gone.
Instead of spending ten minutes on a formula reference site, you describe what you want: "I need a formula that looks up a customer name in column A and returns the matching sales value from column D in a different sheet." You get the formula, usually with an explanation of how it works.
This alone will save you a surprising amount of time.
4. Summarizing and Communicating Results
This might be the most underrated use case. You've done the analysis. The numbers make sense to you. Now you need to write it up for a stakeholder who doesn't want to hear about confidence intervals.
You can paste your key findings into an AI tool and ask it to help you write a clear, non-technical summary. Or ask it to help you frame an insight as a recommendation rather than just a data point. Or help you anticipate the questions your manager is likely to ask so you can prepare your answers.
Data storytelling is a skill, and AI can be a useful thinking partner as you develop it.
5. Learning New Tools on the Job
At some point in your career, you will get a tool that you have never used before, but you will be told to "figure it out" by yourself. For instance, it could be a new BI solution, an unknown database, or some Python module that is not normally part of your toolkit.
On the other hand, AI-based solutions really excel when it comes to fast learning. You can ask a brief explanation of a certain concept, using EL5 before your prompt to ask questions and tell it to break it down to the barest minimum and even get examples of code that fit your case perfectly.
How to Use AI Tools Well (So You Don't Get Burned)
There's a right way and a wrong way to bring AI into your workflow.
Be specific. The more context you give, the better the output. "Write me a SQL query" is going to get you something generic. "Write me a SQL query that calculates month-over-month revenue growth per product category, using a sales table with columns: order_date, product_category, and revenue" is going to get you something actually useful.
Always review the output. AI tools can produce code that looks right but isn't. They can miss edge cases, make assumptions about your data, or confidently give you something that's subtly wrong. Run it, test it, understand it. Don't blindly paste into production.
Don't outsource your thinking. The analysis is essential i.e. what the numbers mean, what decision they support, what context matters and that's yours. AI can help you execute, but the judgment has to come from you. That's where your value as an analyst actually lives.
Use it as a rubber duck. Sometimes I describe a problem to Claude not because I need it to solve it, but because articulating the problem clearly helps me figure it out myself. It's a surprisingly effective thinking tool, not just a code generator.
A Quick Word on Privacy
One thing worth flagging is to be careful about what data you share with these tools. Most AI chat interfaces are not designed to handle sensitive or proprietary data. Don't paste customer information, confidential business metrics, or anything that would make your legal or compliance team unhappy.
Work with anonymized samples, synthetic data, or describe the structure of your data without sharing the actual values. Most of the time, you don't need the real data to get a useful answer anyway.
AI tools aren't going to replace good data analysts. What they will do is aid, if you use them well. It makes you a faster, less frustrated version of yourself. The hours you used to spend on syntax errors and formatting scripts can go toward actual analysis. The cognitive load of remembering every function name drops. The painful parts of the job get a little less painful.
You still need to know your stuff. You still need to understand the business, ask the right questions, and interpret the results with good judgment. But having a capable AI tool in your workflow is starting to feel less like a novelty and more like just... part of the job.
Give it a proper try if you haven't. Not just for one query but actually use it your work. You might be surprised how much it changes things.
Have you started using AI in your data workflow? What's been the most useful application for you? I would love to have your opinion on this. Take this 1-minute job-readiness test to assess your preparedness for the job market.
