Quick Analysis Tools in Excel, and When AI Should Do the Rest

Select a range of numbers in Excel and a small icon appears at the bottom-right corner of the selection — that is Quick Analysis, and most of what people search for under this name is simply not knowing that icon exists. It has been in Excel since 2013, does five specific things well, and is genuinely faster than the ribbon for all five. It is also not the same tool as an AI assistant, and mixing the two up is the more interesting question.

Where it is, and what it does

Select a range of cells with numbers in it. The Quick Analysis icon appears at the bottom-right corner automatically — no ribbon tab, no menu. Click it, or press Ctrl+Q, and five tabs appear:

• —Formatting — conditional formatting presets: colour scales, data bars, icon sets, and highlighting rules like "greater than" applied with one click instead of built manually.
• —Charts — a set of chart types Excel has pre-selected as likely fits for the shape of your data, previewed live as you hover.
• —Totals — running sum, average, count, percentage of total, and a running total, inserted as an actual formula row rather than a static number.
• —Tables — converts the range to a structured Excel Table, or opens it as a PivotTable in one step.
• —Sparklines — a tiny in-cell trend line per row, useful for a quick "is this going up or down" scan across many rows at once.

That is the entire feature. It does not analyse anything in the sense of drawing a conclusion — it applies a formatting or aggregation template to a selection, quickly. The name promises more than the tool delivers, which is exactly why so many searches for it are really searches for something else: a tool that looks at the data and tells you what it means.

Where AI actually picks up

That gap — between "apply a template" and "tell me what this data means" — is where an AI tool genuinely adds something Quick Analysis cannot. Quick Analysis will happily give you a total or a colour scale on a column of numbers; it will not tell you that the total looks unusually low for a Tuesday, or that three rows are duplicates, or which two columns are worth charting together. Reading a plain-language question about the data and identifying what is actually worth looking at is a different task, and it is the one how to create a report in Excel covers for turning that into an actual first-draft writeup.

A worked example

Weak: "Analyse this data."

Better: "Here is a table of monthly sales by region for the last two years. Identify the two regions with the largest month-over-month swings, and tell me in which months those swings happened. Do not calculate any new totals — just point me to the rows and columns to look at, and I will verify the numbers myself in the sheet."

The second version asks the model to do what it is actually good at — reading a description and pointing at a pattern worth investigating — and explicitly keeps the arithmetic in the spreadsheet, where a formula either is or is not correct, rather than in a paragraph you would have to take on trust. How to write a prompt that works on the first try covers the same principle for any request: separate what you are asking the model to find from what you are asking it to calculate. Both OpenAI and Anthropic give the same guidance in their own prompting documentation — state the task and the constraint together, rather than a bare instruction and hoping the model infers the boundary.

Where this goes wrong

Asking an AI tool to do the arithmetic itself — sum a column, compute a percentage change — is the one place it should not replace Quick Analysis or a real formula. Language models are not calculators; they predict plausible-looking text, and a plausible-looking number is not the same as a correct one. This is a documented property of how these models work, not a rare glitch, and a wrong total in a spreadsheet is exactly the kind of error that looks fine until someone downstream acts on it. Quick Analysis's Totals tab inserts a real formula. An AI-generated number in a chat window is a guess wearing a formula's clothing until you check it.

The reverse mistake is asking Quick Analysis to do a job it was never built for — reading a written question about a pattern in your data. It has no language interface at all; it only reacts to what is already selected, which is why searches for "quick analysis tools excel" so often lead people toward an AI feature instead once they realise what they actually wanted.

Let Excel do the arithmetic. Let AI point at what is worth looking at. Keep them in that order.

Checks before you trust either one

1. 01If a Quick Analysis total or chart looks off, check the selected range first — the single most common cause is a selection that missed a row or included a header by mistake.
2. 02If an AI tool flags a pattern in your data, verify it against the actual cells before repeating it in a meeting — checking an AI answer when you are not the expert is the general habit, and a spreadsheet claim is exactly the kind of thing that takes thirty seconds to confirm and does real damage if wrong.
3. 03Look at the raw numbers behind any reference before trusting a formula built on it — the formula bar shows exactly what a cell is actually calculating, which is worth checking whether the formula came from you, a colleague, or an AI suggestion.
4. 04For anything that goes beyond simple totals and charts into genuine optimisation — allocating a budget or a quantity across options — that is Solver's job, not Quick Analysis's, and AI is best used there for setting the problem up rather than solving it.

What to do Monday

Select a range you work with regularly and press Ctrl+Q once, just to see what the five tabs actually offer for your specific data — most people who have used Excel for years have never opened it. Then, separately, take one recurring question you currently answer by scanning a sheet by eye — "which region is off this month" — and ask an AI tool to point you to the rows worth checking, keeping every actual number in the spreadsheet's own formulas. Find the repetitive part is the general test for deciding whether that particular scan is worth automating further, or whether the two-minute manual version is already fine. If you do wire an AI step into a recurring spreadsheet task, decide in advance what you would check to notice it silently broke — the same monitoring principle behind NIST's AI Risk Management Framework, applied to a report instead of a whole system.

This kind of spreadsheet work is exactly where the routine part is shrinking and the judgement about what to check is not — Stanford GSB has written directly about that shift for the roles that live in spreadsheets all day. Coursium teaches the practical version of that judgement — asking the right question of a tool and checking what comes back. Stay ahead of AI by learning the tools on your phone.