Ask AI Questions About a Small Table of Data

You do not need a large dataset to practise using AI with data. A small fictional table is often a better starting point because you can check the results yourself.

The key is to explain what each row and column represents. Without that explanation, an assistant may answer a different question from the one you intended.

Start with a table simple enough to understand without AI.

Use a clear example

Imagine these four orders:

order_id	city	total
101	Nairobi	500
102	Mombasa	800
103	Nairobi	300
104	Kisumu	400

Each row is one order. The order_id identifies the order, city records the delivery city, and total is the order value in the same currency for every row.

Those definitions matter. Two orders from Nairobi do not necessarily mean two different customers. The table has no customer identifier.

Ask one precise question first

Try this prompt:

The table below contains four fictional orders. Each row is one order, and all totals use the same currency. Calculate the combined value of orders delivered to Nairobi. Show the rows you included and the calculation. Use only the supplied table.

The expected result is 500 + 300 = 800. Asking the assistant to show the included rows makes the answer easier to inspect.

You can then ask for the overall total, number of orders or totals grouped by city. Keep the questions separate while learning.

Check calculations independently

The overall total is:

500 + 800 + 300 + 400 = 2,000

There are four orders. The average order value is 2,000 ÷ 4 = 500.

If an assistant returns something different, check which rows it included and what it understood by “average.” Do not accept a result just because it is displayed in a neat table.

For larger or important datasets, use appropriate spreadsheet or database calculations as part of your verification. A conversational explanation alone is not enough.

Keep conclusions within the data

This table does not show profit, customer satisfaction, delivery speed or a monthly trend. An assistant cannot establish those facts from these four columns and rows.

It also cannot explain why Nairobi’s total equals Mombasa’s total. You would need more information to investigate the reason.

Ask the assistant to separate what the table shows from questions that would require additional data. This is often more useful than asking it to produce as many “insights” as possible.

Practise with safe information

Use fictional data first. Before uploading a real table, consider whether it contains private names, account details or confidential business information. Follow the rules that apply to the data and the tool you are using.

Try it yourself

Ask for the city with the largest combined order value. Require the assistant to show all city totals and identify any ties.

Check your answer

Nairobi totals 800, Mombasa totals 800 and Kisumu totals 400. Nairobi and Mombasa tie for the largest combined value. An answer naming only one of them misses an important detail.

Good data questions produce answers you can trace back to the rows, rather than impressive claims that go beyond the evidence.