AI for Data & Spreadsheets
6.3 — AI for Data and Spreadsheets
What You'll Learn
You don't need to be a data analyst to get useful insights from data. AI has made it possible to have a plain-English conversation with your spreadsheets — asking questions, spotting trends, and generating summaries without writing a single formula.
In this lesson, you'll learn how to use AI to work with data: from explaining what your spreadsheet is doing, to analysing customer feedback, to spotting trends in survey results. We'll walk through real examples so you can see exactly what's possible.
Explaining Spreadsheets in Plain English
If you've ever inherited a spreadsheet built by someone else — full of nested formulas, colour-coded columns, and unexplained abbreviations — you know the frustration. AI can help.
What to do: Copy a portion of the spreadsheet (paste it as text, or screenshot it) and describe what you're seeing. Ask AI to explain what formulas are doing, what the columns mean in context, or how the data is structured.
Example prompt:
"I've inherited a spreadsheet with these columns: [paste columns]. There's a formula in column F that reads =IF(D2>0, C2/D2, "N/A"). Can you explain what this is calculating and when it returns N/A?"
AI response (simplified): The formula is calculating the ratio of column C to column D — probably a conversion rate, margin percentage, or similar ratio. It returns "N/A" when column D is zero, because dividing by zero isn't possible.
This takes seconds and saves the embarrassment of having to ask a colleague or Google the formula syntax.
Analysing Customer Feedback
Customer feedback — from surveys, support tickets, review sites, or emails — is one of the most valuable and most neglected data sources in most businesses. It's text-heavy, which makes it hard to analyse with traditional tools. AI is very good at this.
What you can do:
- Paste in 50–200 customer comments and ask AI to identify the most common themes
- Ask it to separate positive from negative feedback
- Ask it to find the most urgent issues mentioned
- Ask it to count how often specific topics appear
Real worked example:
Suppose you've collected 80 responses to the question: "What's the one thing you'd most like us to improve?"
Prompt:
"Here are 80 customer responses to the question 'What would you most like us to improve?' [paste responses]. Please: 1) identify the top 5 most common themes, 2) give me 2 representative quotes for each theme, 3) note which themes are most emotionally intense."
Output (example):
Top themes: 1) Onboarding complexity (23 mentions) — "I had no idea where to start"; "The setup took me three days when I expected three hours". 2) Pricing transparency (18 mentions)...
In 30 seconds you have a qualitative analysis that would have taken an analyst hours to produce manually.
Key takeaway: AI doesn't replace customer research — it makes it dramatically faster to surface what customers are actually saying.
Using ChatGPT's Data Analysis Feature
ChatGPT (on the paid Plus plan) has a feature called Advanced Data Analysis (previously Code Interpreter) that can directly read uploaded spreadsheet files and do real computation on them.
How it works:
- Go to ChatGPT and start a new conversation
- Click the paperclip icon and upload your CSV or Excel file
- Ask questions in plain English
What it can do:
- Calculate totals, averages, medians, and percentages
- Create charts and visualisations
- Filter and sort data
- Identify outliers or anomalies
- Perform regression analysis (yes, really)
- Compare categories or time periods
Example prompts that work well:
"Show me which product category had the highest revenue growth between Q1 and Q3."
"Create a bar chart comparing customer satisfaction scores across our four regions."
"Which 10 customers have the highest lifetime value? Show me as a table."
"Are there any months where sales dropped significantly? What might explain the pattern?"
ChatGPT will write and run code under the hood, then show you the results. You never have to see the code if you don't want to.
Important limitation: The file stays within that session. ChatGPT doesn't store or learn from your data, but you should still avoid uploading sensitive personal data or confidential financials to external AI tools unless your organisation has a secure enterprise agreement in place.
Summarising Survey Results
If you run internal surveys — staff engagement, project retrospectives, product feedback — AI can turn the raw results into a readable summary report in minutes.
Worked example:
You ran a team survey with 40 responses. You export it as a CSV. You upload it to ChatGPT with Advanced Data Analysis.
Prompts in sequence:
- "Summarise the overall sentiment from the open-text responses in the 'comments' column."
- "For the rating questions (columns D–H), calculate the average score for each and rank them from highest to lowest."
- "Are there any statistically significant differences in ratings between the two departments in column B?"
- "Write me a 300-word summary of the survey findings suitable for sharing with the leadership team."
Each prompt builds on the last. By the end, you have a structured summary that would otherwise take hours to write.
Spotting Trends
One of AI's most useful capabilities with data is identifying patterns humans tend to miss — particularly across time.
Try these prompts with your own data:
- "Look at these monthly figures over 18 months. Are there any seasonal patterns?"
- "Which metric has been most consistently improving? Which has been most volatile?"
- "If the current trend continues, what would this number look like in 6 months?"
That last prompt is not a forecast — treat it as indicative, not predictive. But it's a useful conversation starter for planning discussions.
Practical Tips
- Start with clean data. AI can handle messy data, but you'll get better results if columns are labelled clearly and there are no rogue merged cells.
- Ask follow-up questions. If the first answer is vague, push: "Can you break that down by month?" or "Which 5 rows are driving that pattern?"
- Ask AI to check its own work. "Does that calculation seem right given the data?" sometimes catches errors.
- Export and document. Copy AI-generated analysis into your own documents and note the date and source. Don't rely on the AI chat log as a record.
Key takeaway: You don't need to be a data analyst to get genuine insights from data. With AI, the skill is asking good questions — not running complex formulas.
What to Try This Week
Find a spreadsheet you've been meaning to dig into — customer data, sales figures, survey results, anything with more than 50 rows. Either paste a section as text into ChatGPT/Claude and ask for a summary, or upload the file using ChatGPT's Advanced Data Analysis feature. Ask three genuinely useful questions. See what you learn.