🤖 Part 10: Data & Spreadsheets

Spreadsheets are where a lot of us feel out of our depth — formulas that won't cooperate, a table of numbers that must be hiding a story. An assistant is a wonderful partner here: it explains, summarizes, and writes formulas in plain English. But it's also a text predictor, not a calculator, so this is the one part where the "verify what matters" rule earns its keep the most.

📚 What You'll Learn

By the end of this part, you'll be able to:

Paste a small table or CSV and get summaries, trends, and plain-English insights from it
Generate and understand formulas for both Excel and Google Sheets — and have each one explained
Clean up messy data and generate realistic sample data for testing or practice
Know the arithmetic caution — why a text predictor can miscount, and when to switch on a real analysis or code tool for exact math

In This Part

Pasting a Table and Asking What It Means

You don't need to upload a whole spreadsheet to get value — for a small table, the fastest path is to copy the rows and paste them straight into the chat. Copy a range out of Excel or Google Sheets and it usually pastes as neat rows; or paste comma-separated values (CSV). Then ask your question. The assistant reads the table and talks about it like a colleague looking over your shoulder.

The most useful requests fall into a few buckets:

You want…	Ask something like…
A summary	"Summarize this table in plain English — what are the three most important things it shows?"
Trends & patterns	"Which month grew fastest? Is there a trend over time, and where's the outlier?"
Insights for a decision	"If I had to cut one product from this list, which does the data suggest, and why?"
Reshaping	"Turn this into a tidy table sorted by revenue, highest first, with a total row."

This is genuinely where an assistant earns its keep for non-experts: it puts words to numbers. A column of figures that meant nothing becomes "sales dipped every January and peaked in summer — your business is seasonal." That plain-English read is often exactly the insight you were missing, and it costs you one paste and one question.

🧠 Mindset

Treat the assistant as a sharp analyst giving you a first read, not a final audit. Its narrative about your data — what's rising, what's odd, what to look at — is usually insightful and always worth having. Whether a specific number in that narrative is exactly right is a separate question, and one you'll learn to check in a moment.

Formulas for Excel and Google Sheets

Here's a place an assistant is close to a superpower: writing spreadsheet formulas. Describe what you want in plain words, name your tool, and it produces the formula — and, if you ask, explains every piece so you actually learn it instead of just pasting a black box.

The magic phrase is "and explain it". Compare:

"Write an Excel formula to add up column B only where column A says 'Paid'." → you get =SUMIF(A:A,"Paid",B:B).
"…and explain each part so I understand it." → you also learn that SUMIF checks a range for a condition and totals the matching values, what each argument means, and how to adapt it.

Because Excel and Google Sheets mostly overlap but occasionally differ, always say which one you're using — the assistant will match the right function names and quirks. And when a formula doesn't work, paste it back with the error and ask why; debugging formulas is one of its most reliable tricks.

✅ Tip — ask for the plain-English version too

Request "the formula, then a sentence describing what it does, then how I'd change it for a different column." You walk away with something that works and something you understand — so next time you can tweak it yourself. That's the difference between borrowing an answer and building a skill.

⚠️ Important: A formula the assistant writes is logic it's confident about — but it's built for your layout, which it can only guess at from your description. Check that the cell ranges, column letters, and sheet names actually match your file before you trust the result. A perfect VLOOKUP pointed at the wrong column returns a confidently wrong number.

The Arithmetic Caution

Now the most important idea in this part, and it flows straight from Part 1. An assistant predicts language. It does not, by default, calculate — it produces the text that a correct answer usually looks like. For simple sums that's often right. But ask it to add a long column, average a big list, or do multi-step math in its head, and it can quietly land on a number that's close but wrong, stated with total confidence.

graph TD A["🔢 You ask for a number
sum, average, count, %"] --> B{"Must it be exact?"} B -->|"No — a rough read is fine"| C["💬 Plain prediction
usually close, sometimes off"] B -->|"Yes — it will be acted on"| D["🧮 Turn on data-analysis /
code mode ⟨if available⟩"] D --> E["✅ Real computation
you still sanity-check"] C --> F["🔎 Sanity-check the figure
before you rely on it"]

Here's the good news, and the reason this isn't a dead end: many assistants now offer a data-analysis or code mode that writes and runs real code to crunch the numbers — actual computation, not a language guess. When that mode is on, the math is trustworthy in a way plain chat isn't. Availability and naming vary by product and tier (some call it advanced data analysis, code interpreter, or similar), so we won't promise it's in your specific plan — but when exact math matters, look for it, and turn it on.

⚠️ Watch Out — the confident wrong total

The dangerous case isn't a number that's obviously absurd — you'd catch that. It's a total that's plausible and slightly off: $12,840 instead of $12,480. Nothing flags it. So for any figure you'll report, bill, or decide on, either use the data-analysis mode or verify it in the spreadsheet itself. Never let a chatted-out number be the last word on money, dosages, or deadlines.

📖 Definition

Data-analysis / code mode: a feature where the assistant writes real program code and executes it to compute an answer, then reports the result. Because a computer actually does the arithmetic, the numbers are reliable — a genuine calculator behind the chat, rather than a prediction of what the answer should be.

Cleaning Messy Data & Making Sample Data

Two more everyday wins round out data work — and both play to the assistant's strengths as a language tool, so they're safe and fast.

Cleaning and reformatting messy data. Real-world data is a mess: names in mixed capitalization, dates written five different ways, extra spaces, "N/A" scattered through number columns. Paste the messy rows and describe the tidy version you want:

"Standardize these dates to YYYY-MM-DD and fix the inconsistent capitalization in the names column."
"Split this 'Full Name' column into 'First' and 'Last', and trim the extra spaces."
"Flag any rows that look like duplicates or have obviously wrong values, but don't delete them."

Generating realistic sample data. Need a test table, a demo, or practice data that looks real without using anyone's actual information? Just ask: "Make a 12-row sample table of fictional customers with name, city, signup date, and monthly spend — realistic but entirely made up." It's perfect for trying a formula, building a template, or learning, without touching sensitive real records.

✅ Tip — clean small, then check, then scale

When cleaning data, do a small batch first and eyeball the result against the original — did it change anything it shouldn't have? Once you trust the pattern, ask it to apply the same rules to the rest, or to write a formula or steps you can run yourself in the spreadsheet. That keeps you in control of your real data rather than handing it over wholesale.

⚠️ Important: Cleaning is a transformation, and transformations can quietly drop or alter rows. Always compare the count and a few sample rows before and after. And if the data is real and sensitive, remember Part 13 on privacy — practicing with generated sample data first is often the smarter move.

🛠️ How To: Summarize a Table and Get a Working Formula

What you'll do: paste a small table, get a plain-English summary of what it shows, and walk away with a spreadsheet formula you understand — the everyday data workflow in one pass.

Step by step

Copy a small, non-sensitive table. A dozen rows from a personal budget, a sales sheet, or the sample data the assistant can generate for you. Avoid anything private for now (Part 13).
Paste it into the chat and ask: "Here's my table. Summarize what it shows in plain English, and point out any trend or outlier."
Ask for a formula, naming your tool: "I use Google Sheets. Give me a formula to total the 'Amount' column only for rows where 'Category' is 'Groceries' — and explain each part."
Paste the formula into your spreadsheet, fixing the column letters and ranges to match your real layout. Confirm it returns a sensible number.
Sanity-check the math. Compare the assistant's spoken total (if it gave one) against what the spreadsheet computes. Trust the spreadsheet. If exact figures matter and a data-analysis mode is available, use it.

💡 Tip

Keep it all in one conversation and iterate: "now break that total out by month," "now add a formula for the average," "now explain why my version returned an error." The assistant remembers your table and tool, so each follow-up gets faster — and you're learning the formulas as you go, not just collecting them.

Best Practices

✅ Do's

Name your tool — Excel or Google Sheets — so formulas match its function names and quirks.
Ask it to explain every formula so you learn it and can adapt it, instead of pasting a black box.
Use the data-analysis / code mode for exact math when it's available — real computation beats a language guess.
Practice with generated sample data before touching real, sensitive records.

❌ Don'ts

Don't trust a chatted-out total for anything you'll bill, report, or act on — verify it in the spreadsheet or with the code mode.
Don't paste a formula without checking its ranges — a correct formula pointed at the wrong column is confidently wrong.
Don't assume cleaning kept every row — compare counts and sample rows before and after.
Don't upload sensitive spreadsheets until you've read Part 13 on privacy.

💡 Pro Tips

For a big or important calculation, ask "show me the formula I can run myself" rather than "give me the answer" — then your spreadsheet does the exact math, not the chat.
When a number looks surprising, ask the assistant to walk through the calculation step by step. It often catches its own slip — and you learn where the risk was.

📓 Learning Journal

Keep a journal as you work through this guide — digital or paper. After each part, jot down:

Key ideas you learned
Things that clicked for you
Questions or confusion points to revisit
Ideas you want to try
Your progress and how you feel about it

✍️ This part's prompt: Take a small table from your own life — a budget, a list of expenses, a log of anything you track — and ask the assistant to summarize it and write one formula you've always wished you knew. Then deliberately test the arithmetic caution: ask it to total a column in the chat, and check that total against your spreadsheet. Was it right? Write down where you'd trust its words and where you'll always reach for real computation.

📝 Part Summary

🎓 Key Takeaways

Paste a small table or CSV straight into the chat to get plain-English summaries, trends, and insights — a fast first read of your numbers.
Describe what you want and name your tool to get Excel or Google Sheets formulas — and always ask it to explain each part so you learn it.
The arithmetic caution: an assistant predicts language and can miscount; for math that must be exact, use a data-analysis / code mode (varies by product and tier) or your spreadsheet.
It's great at cleaning messy data and generating realistic sample data — just compare rows before and after, and sanity-check every number that matters.

🎉 What You Can Now Do

You can hand an assistant a table and get back the story hidden in it, then walk away with a formula you actually understand and can adapt. You can turn a mess of inconsistent data into something tidy, and conjure realistic practice data out of thin air. Most valuable of all, you know the one line you must never cross: the assistant is a brilliant explainer of numbers and a shaky calculator of them, so you reach for real computation whenever the figure truly counts. That single instinct will save you from the confident-wrong-total trap for good.

❓ Common Questions

Why did it get a simple total wrong?

Because in plain chat it isn't calculating — it's predicting the text of an answer, and for longer or multi-step math that prediction can land near the right number but not on it. The fix is to have it produce a formula you run in your spreadsheet, or to switch on a data-analysis / code mode that does real computation.

Should I paste my data in, or upload the spreadsheet file?

For a small table, pasting is quickest and clearest. For a large file, uploading may work (subject to the size and truncation limits from Part 8, and whether your tool supports spreadsheet files). Either way, keep sensitive data out until Part 13, and lean on the code mode when the numbers must be exact.

Do all the assistants have a data-analysis mode?

Not identically. Several offer some form of it, but names, availability, and whether it's on the free tier vary by product and change over time, so we won't promise it's in your plan. When exact math matters, look for a feature that says it runs code or analyzes data, and check the tool's official help page for current details.

🔭 Up Next

In Part 11: Hallucinations — Confident, and Sometimes Wrong, we'll face the guide's central caution head-on: why assistants invent facts, fake citations, and plausible-but-false details, how to spot the warning signs, and how to work so a confident mistake never slips through unchecked.

📚 Additional Resources

🌟 Keep Going

Spreadsheets just got a lot friendlier — you've got a partner that explains your data and writes your formulas in plain words. Keep the one golden habit close: brilliant with language, careful with numbers, real computation when it counts. Next up, we look the assistant's confident mistakes right in the eye. 🤖