Essential Hard Skills: Spreadsheets, Slides, and AI

At one point, being "good at Excel and PowerPoint" was a real competitive edge.

That sounds ridiculous now. But it wasn't then. The people who learned spreadsheets before their peers got handed the interesting problems.

Today these skills are table stakes. Nobody gets promoted for knowing what a pivot table is. But plenty of people quietly lose trust because their numbers don't add up, their file breaks when someone sorts it, or their slides take ten minutes to explain.

That's the real stakes. Hard skills aren't about looking technical. They're about whether people can trust your work without redoing it.

And now there's a third tool on the list. AI assistants can write a formula, clean a messy export or draft a slide outline in seconds. That's a genuine speed advantage. It's also a fast way to send a wrong number to your boss with total confidence.

This module covers the handful of spreadsheet moves that do most of the work, how to build a slide that makes one point, and how to use AI to go faster while still owning the accuracy.

The short version

• Clean data first. One header row, one row per record, one type of thing per column. Most spreadsheet pain comes from skipping this.
• Learn four moves well: filters and sorting, SUMIFS/COUNTIFS, lookups (XLOOKUP), and pivot tables.
• One slide, one point. The headline is the takeaway, written as a full sentence. Everything else on the slide proves it.
• Use AI to write formulas, explain them, clean data and draft outlines. Never send a number you haven't checked.
• Your name is on the work, not the tool's.

The spreadsheet moves that do most of the work

You don't need macros or advanced modelling to be the person people trust with data. You need a solid base, built in this order:

Clean data

Before any formula, set the data up so it can be used:

• One header row. Short, clear column names. No blank header cells.
• One row per record. One order, one customer, one ticket per row.
• One kind of thing per column. Dates in the date column. Don't mix "N/A" into a number column.
• No merged cells, no blank rows in the middle. They break sorting, filtering and pivots.
• Raw data on its own tab. Do your analysis on a separate tab so you never overwrite the source.
• Consistent formats. "Calgary", "calgary" and "Calgary " (with a space) are three different cities to a spreadsheet. TRIM and PROPER fix most of this.

Turn the range into a table (Ctrl+T in Excel, or a filter view in Google Sheets). Formulas stay readable and new rows get picked up automatically.

SUMIFS and COUNTIFS

These answer "how much" and "how many" with conditions. Total sales for the West region in March. Number of tickets still open over seven days. Once you can write these, you stop hand-counting forever.

Lookups

A lookup pulls a value from one table into another using a shared key, like an employee ID or order number. Use XLOOKUP (Excel and Google Sheets both support it). It's simpler and safer than the older VLOOKUP, which breaks if someone inserts a column. If you inherit VLOOKUPs, learn to read them, but write new ones as XLOOKUP.

The classic bug: the lookup returns nothing because one side has a trailing space or a number stored as text. Clean the key column first.

Pivot tables

A pivot table summarizes a big table in a few clicks. Drag a field into rows, another into values, and you can see totals by region, by month, by rep, without writing a formula. It's the fastest way to answer the question your manager asks in a meeting.

A slide that makes one point

Most slides fail because they try to do three jobs. A chart, a table, four bullets and a title like "Q3 Results." The audience has to work out what you think.

Flip it. Write the point first.

The headline is the takeaway. Not "Q3 Customer Complaints." Instead: "Complaints fell 20% after we changed the delivery window." If someone only reads your headline, they should get the point.

One piece of evidence. One chart that proves the headline. Highlight the bar or line that matters and grey out the rest.

The headline test. Read only your slide titles in order. If they tell the story, the deck works. If they read like a table of contents, rewrite them.

Keep a simple order for most decks: the situation, the problem, what you recommend, and what you need from the room.

Using AI to go faster (while still owning accuracy)

AI assistants are now built into Excel, Google Sheets and most slide tools, and general assistants like ChatGPT, Claude and Gemini can do much of the same work. Used well, they save real hours. Used badly, they produce confident nonsense.

Good jobs for AI:

• Writing a formula from a plain description: "Sum column D where column B is West and column C is March."
• Explaining a formula you inherited, step by step.
• Cleaning messy text: splitting names, fixing date formats, standardizing city names.
• Suggesting how to structure a pivot or which chart fits the data.
• Drafting a deck outline or five headline options for a slide.

Rules for owning the output:

1. Check the numbers. Recalculate the grand total a second way. Spot-check three rows by hand. If the AI summarized a file, open the file.
2. Make it explain itself. If you can't explain the formula to your manager, don't ship it.
3. Mind the data. Don't paste customer, employee or financial data into a tool your company hasn't approved. Ask what's allowed. It's a short question that can save your job.
4. Write the point yourself. AI is good at polish. The judgment about what matters is yours.

Key idea: AI doesn't replace knowing how spreadsheets work. It amplifies it. If you understand the basics, you get much faster. If you don't, you just make mistakes faster.

What it looks like in practice

A composite example, not a real person.

Marcus is an operations analyst, a year into his first full-time job. Every Monday he builds a report on late deliveries by region. It takes him most of the morning, and twice this quarter his manager caught a total that didn't match.

He rebuilds it in an afternoon. First, he moves the raw export to its own tab and turns it into a table. He asks an AI assistant to write a formula that trims spaces and standardizes region names, then has it explain the formula back so he knows what it does. He replaces his hand-counted totals with COUNTIFS and builds one pivot table: late deliveries by region by week.

Then he checks. The pivot's grand total matches the row count of the raw data. He spot-checks three deliveries against the source system. One region is off, and he finds why: a new depot code the cleanup formula didn't know about. He fixes it.

His Monday slide goes from a wall of numbers to one headline: "Late deliveries in the East doubled since the depot change." One bar chart, East highlighted.

The report now takes him under an hour. His manager stops double-checking his totals. That's the real win.

Common traps

• Analysing dirty data. Every later step inherits the mess. Clean first, even when you're in a hurry.
• Hard-typing numbers into formulas. Someone updates the data and your total silently goes wrong. Reference cells instead.
• Slide titles that are labels. "Q3 Results" makes the audience do the thinking. Write the takeaway.
• Too much on one slide. If you need to say "this is a busy slide," split it.
• Trusting AI output because it looks right. Polished isn't the same as correct. Check totals and spot-check rows.
• Pasting sensitive data into any AI tool. Check your company's policy first.

Do this this week

Take one recurring report or file you work on and level it up.

1. Move the raw data to its own tab, turn it into a table, and fix the formatting issues from the clean-data list.
2. Replace one manual step with SUMIFS, COUNTIFS, XLOOKUP or a pivot table. Use an AI assistant to help write it, and make it explain the formula back.
3. Check the result two ways: grand total and three spot-checked rows.
4. Build one slide with a full-sentence headline and one chart. Show it to someone and ask, "What's the point of this slide?" If they say your headline, you're done.

Your tool

CopyBEFORE YOU HIT SEND: SPREADSHEET + SLIDE + AI CHECKLIST

DATA
[ ] One header row, one row per record, one type per column
[ ] No merged cells, no blank rows in the middle
[ ] Raw data on its own tab; analysis on another
[ ] Text cleaned (TRIM, consistent spelling, dates as dates)

NUMBERS
[ ] No hard-typed numbers inside formulas
[ ] Grand total checked a second way
[ ] 3 rows spot-checked against the source
[ ] I can explain every formula out loud

SLIDE
[ ] Headline is a full sentence stating the takeaway
[ ] One chart or number proves it; key part highlighted
[ ] Units, dates and source labelled
[ ] Extras moved to the appendix
[ ] Reading only the headlines tells the story

AI USE
[ ] Tool is approved for this kind of data
[ ] I asked it to explain anything I didn't fully follow
[ ] I edited the wording so it sounds like me
[ ] I'd defend every number if my manager asked

FORMULA CHEAT SHEET
=SUMIFS(sum_range, criteria_range1, criteria1, ...)
=COUNTIFS(criteria_range1, criteria1, ...)
=XLOOKUP(lookup_value, lookup_array, return_array, "Not found")
=TRIM(A2)          removes extra spaces
=PROPER(A2)        Fixes Capitalization
=IFERROR(formula, "")   hides errors, but check why first

PROMPTS THAT WORK
"Write a formula for [Excel/Sheets] that [plain description].
 My data: column A = ..., column B = ..."
"Explain this formula step by step: [paste]"
"Give me 5 headline options that state this takeaway: [your point]"

Talk about it

1. What's one file or report at work that people don't fully trust? What would it take to fix it?
2. Where has an AI tool saved you real time, and where has it burned you?
3. Pull up a deck you've built recently. Do the slide titles tell the story on their own?

— Evan