Skip to content
BENIBIX Learn

AI Coding and Data Analysis

Level 3 · Assistants at work

How an AI coding tool suggests and fixes code. How an AI data analyst runs code on your sheet to answer you, so the sums are not guesses.

Course contentsLesson 7 of 13

After this lesson you can

  • Say what code completion does
  • Explain why a data analyst runs code and does not guess
  • Check the work of a coding or data assistant

10 mincovers 5 glossary entries

Worth reading first: Writing and Meeting Assistants

How does AI help with code and with data?

An AI coding assistant helps people write, read and fix code. GitHub Copilot, one of the first, went on sale on 21 June 2022 after a year of testing. An AI data analyst answers questions about your data. It writes a short program, runs it on your spreadsheet and shows the answer, often as a chart.

The simplest kind of coding help is code completion. As you type, the tool shows the next few lines in grey. You press a key to take them. Bigger tools can answer questions about a whole project and explain an error. They can change many files and run the tests to see what broke.

The person still decides what goes in. A coding assistant writes code that looks right, and most of it is. Some of it calls things that do not exist, or misses a case nobody told it about. Read what it wrote, run it and test it, the way you would check a new team member's work.

An AI data analyst works on the numbers you give it. You upload a spreadsheet and ask in plain words, such as which month sold the most. It writes a short program, runs it on your file and reads the result. You get a table or a chart, and you can ask a follow-up.

Running real code is the point. A chatbot that guesses a sum from memory can get it wrong. An analyst that runs the sum on your rows works it out. Look at the code it ran, even if you cannot write code. Check that it used the right column and the right months.

Both tools need clean input. A spreadsheet with blank rows, mixed dates or two names for the same shop will give a wrong answer with a neat chart on top. The same goes for code, where a vague request gets a vague program. Say exactly what you want, then check what you got.

In each row the AI does the slow middle step, and a person checks the last one.

Worked example

Emma's sales sheet

Emma uploads a sheet of 1,200 sales and asks which of her 4 shops grew most this year.

The analyst runs code, draws a chart and says the city shop grew 18%. Emma opens the code and sees it counted refunds as sales. She asks it to drop them, and the answer becomes 11%.

The words for this lesson5

Check yourself

Why is an AI data analyst better at sums than a plain chatbot?

Finished this lesson?

Marking it complete also marks its 5 glossary entries learned, which updates your roadmap and your saved page. 0 are already marked.