Your first AI wins: 5 everyday tasks

When you finish, you can use AI for queries, formulas, cleaning, charts and summaries, and know what to check in each.

Win one: a query from a plain question

Weak prompt: “Write SQL for sales by region.” The assistant has to guess your tables. Strong prompt: “I use PostgreSQL. Table orders has order_id, customer_id, order_date, amount. Table customers has customer_id, region. Write a query for total amount by region for last month, excluding refunds where amount is negative. Explain each join.”

Check three things. First, row counts: does the total match a figure you already trust? Second, the join: could a customer appear twice and inflate the sum? Third, the dates: is “last month” using the right boundaries? Describe your schema every time, using names only, never real rows.

Win two: formulas you can understand

Weak: “Make a formula to get the average.” Strong: “In column D I have order values and in column B the region. Write a spreadsheet formula for the average of D where B equals West, ignoring blank cells. Explain each part.” The explanation matters, because you will maintain this formula.

Test on three rows where you know the answer. Then test an ugly row, such as a blank or text like “n/a.” Many formulas fail quietly on those. Ask the assistant, “What inputs would break this?” and try them.

Win three: plan the cleaning

Weak: “Clean this data.” Strong: “Here are the column names and five invented sample rows. List the checks I should run before analysis: duplicates, inconsistent spellings, impossible dates, text in number columns, and missing values. Give me a way to detect each in Excel or Python.” Now you have a plan, not a mystery.

Run every check yourself and keep a note of what you changed and why. Never let a tool silently drop rows. After cleaning, compare row counts before and after. If 400 rows became 380, you should be able to say why 20 left.

Win four: pick and test a chart

Weak: “Make a nice chart.” Strong: “I want to show a manager that late deliveries rose after we changed carriers. The data has week, carrier and percent late. Suggest two chart types, pick one, and name how each could mislead.” This pushes the tool to think about the viewer.

Check the axes. A bar chart that does not start at zero can exaggerate small differences. Check the labels and the title: does the title state the finding? Remove anything decorative that does not help the reader see the point.

Win five: a summary a manager will read

Weak: “Summarize this.” Strong: “Write a half-page summary of these results for a busy operations manager. Lead with the answer, give three supporting points, state one limitation, and suggest a next step. Use only the numbers I paste.” The last sentence is important.

Compare every number in the summary to your source. Assistants sometimes round, swap or invent figures. If a number is not in your data, delete it. The summary goes out under your name, so you own each claim.

Practice now

Describe a table you use at work using only column names and invented rows. Ask an AI tool for a query on a real question, then run it on a small slice and check the totals.
Ask an AI tool to suggest two chart types for one of your findings. Build one, then list how it could mislead a viewer.

Use any AI assistant you have. Do not paste private, client or patient information into it.

Check yourself

1. What is the main risk when an AI writes a join between two tables?

2. After cleaning data, what simple check helps you trust the result?

3. What should you do with numbers in an AI-written summary?

Ask the AI tutor

Ask anything about this lesson, in any language. The tutor is an AI: it can be wrong, and it never needs your personal details.

Next lesson: Guardrails for data analysts

AI-drafted. Checked by an automated validator and a separate AI reviewer; a person has not read every lesson. Translations and tutor answers are AI-made. Found a mistake? Tell us. AI can be wrong. General information only, not professional advice. Not affiliated with any AI company named. Current University is not accredited.