The signal

You do not need to build an AI app to make yourself more useful at work. Start with something less glamorous: a spreadsheet someone checks by hand every week.

Maybe it tracks unpaid invoices, missing application details or orders waiting for dispatch. An AI assistant can help turn a plain-English rule into a spreadsheet formula. That lowers the barrier for beginners who know what needs to happen but cannot remember the right function.

The valuable skill is not getting AI to produce a formula. It is knowing whether that formula works. A confident-looking answer can quietly label the wrong rows, especially when dates, blanks or inconsistent spelling are involved.

For someone in a first office job or running a small side business, this is an approachable workflow to practise. You can demonstrate a useful result without buying a course or pretending to be a programmer.

How it actually works

Think of AI as a formula drafter, not the person responsible for the spreadsheet. You supply the rule, test the output and decide whether it is safe to use.

Start by writing the rule without spreadsheet language. For example: “Flag an invoice for follow-up only when its due date is before our review date and it is marked unpaid.”

Notice the decisions hiding inside that sentence. Does an invoice due on the review date count? What happens when the due date is missing? Does an empty payment-status cell mean unpaid, or does it mean someone needs to check?

Set those rules before asking for a formula. Otherwise, the assistant may make decisions you never intended.

A simple workflow is:

  • Create a practice sheet with invented information, not a copy of customer or employer records.
  • Tell the assistant which spreadsheet app you use and what each column contains.
  • Ask for one formula and a plain-English explanation of every part.
  • Test it against outcomes you wrote down independently.
  • Keep a human review step before sending messages or changing records.

Do not paste confidential data into a personal AI account. Follow your employer’s approved-tool and data-handling rules. Removing names alone may not make a dataset safe to share.

Real example

An administrative assistant could use ChatGPT to draft an overdue-invoice flag for Microsoft Excel. The first version should be a practice sheet, not a live collections workflow.

Put due dates in column A, payment status in column B, and a fixed review date in F1. Use real spreadsheet date values rather than dates stored as text. Then try this prompt:

“I’m using Microsoft Excel. A2 contains a due date, B2 contains either Paid or Unpaid, and F1 contains the review date. Write a formula for C2 that returns Follow up only when the due date is before the review date and the status is Unpaid. Return Check if the date is blank or not a numeric date value, or if the status is not one of those two labels. Otherwise return a blank. Keep F1 fixed when I copy the formula down. Explain the formula and its assumptions.”

Build six test rows: an overdue unpaid invoice, an overdue paid invoice, a future unpaid invoice, one due on the review date, one missing its date and one missing its status.

Write your expected labels before running the formula. If the results disagree, investigate rather than asking AI to reassure you. A numeric value can also be an implausible date, so passing these checks does not settle every possible problem.

The useful work sample is the sheet plus its rules and tests—not a claim that you “automated finance.”

Your small move today

Spend 15 minutes making that six-row practice sheet and testing one AI-drafted formula. Use an account you already have; no subscription is needed for the exercise.

Save the sheet with one sentence explaining what it checks and what still needs human review. That is a small, concrete piece of evidence that you can use AI carefully, not just quickly.

SOURCES
  1. 1OpenAI — ChatGPT overview
  2. 2OpenAI Help Center — Data analysis with ChatGPT

Drafted with AI from trusted public sources and checked against our editorial rules. Educational content only — not personalized investment advice. Investing has risks and returns are never guaranteed.