Using AI at Work: A Task-by-Task Guide

Using AI at work pays off fastest on the jobs you already know how to do, because you can tell in seconds whether the draft is any good.

This guide walks through four of the most common: writing, research, spreadsheets and customer support. For each one you'll get the pattern that works, a prompt to copy and the check that stops a mistake reaching a customer. It assumes a general chat assistant; if your business software has AI features built in, the same principles apply, but check the vendor's current documentation for what those features do with your data.

Pick the Right First Task

A good first task has three qualities: it repeats every week, you are already good at it, and an error is cheap and easy to spot. A bad first task is rare, high-stakes or something you don't fully understand yourself. The aim of the first month is not to automate your business; it is to learn how the tool behaves on work you can judge.

Using AI for Writing

Writing is where most people feel the time saving first. The pattern: you supply the facts as rough bullet points, the assistant supplies shape and wording, and you edit the result.

1. Write the facts in bullet form: who, what happened, what you want the reader to do.
2. Say who the reader is and paste a past example of your tone.
3. Ask for one version at a set length, then ask for changes.
4. Read it as the customer would. Remove anything you didn't supply.

Turn these notes into a short email to a customer. Keep my facts exactly as given and don't add any new ones.

Reader: a long-time customer who asked about delivery.
Notes:
- their order ships Thursday, not Tuesday
- the delay is on our side
- we'll include the matching cushion covers at no charge

Tone: warm, plain, no exclamation marks. Under 120 words. Sign off with [MY NAME].

Using AI for Research

Treat an assistant as a fast research assistant who has read a lot but cannot be trusted with a citation. It is excellent at telling you what you don't know yet: the terms to search, the questions to ask a supplier, the angles you missed. It is weak as a final source.

• Good: "What questions should I ask three commercial cleaning companies before signing a contract?"
• Good: "Summarise these three supplier quotes I've pasted, in a table, and list what's different."
• Risky: "What is the current minimum wage in my state?" Get that from the official source.

If your tool can search the web, ask it to list its sources, then open them. Check the date on each page. A summary of an outdated page is still outdated. The guide to checking AI output has a step-by-step method.

Using AI for Spreadsheets

Describing what you want in plain English and getting a working formula back is one of the most dependable everyday uses of AI. The safest pattern keeps your data on your machine: ask for the formula, not for the answer.

I use [SPREADSHEET SOFTWARE]. Column A has invoice dates, column B has amounts, column C has the date paid (blank if unpaid).

Write a formula for column D that shows "Overdue" if the invoice is more than 30 days old and unpaid, "Paid" if column C has a date, and stays blank otherwise. Explain what each part of the formula does.

Then test it. Pick four rows where you already know the right result, including an edge case such as a blank date or an invoice dated exactly 30 days ago. If any row is wrong, paste the row and the wrong result back into the chat and ask it to fix the formula.

Using AI for Customer Support

For a small business, the safest and most useful support pattern is not a chatbot talking to customers on its own. It is an assistant helping the person on the desk answer faster and more consistently.

1. Write down the ten questions you get most, with the answers you actually give.
2. Use AI to tidy them into a clear reply library in your tone.
3. When a new message arrives, paste the question (minus personal details) with the relevant library answer and ask for a reply draft.
4. A person reads, edits and sends it.

If you later consider an automated assistant on your website, read AI agents explained first, and be clear with customers that they are talking to software.

Run a Two-Week Trial

Before you make AI part of how the business runs, run a small, honest trial. Pick one task, do it with AI for two weeks and keep a simple log: the task, roughly how long it took, what you had to fix, and if you'd send it as-is. At the end you'll know whether it saves time on that job, and you'll have a prompt worth keeping.

The Bottom Line

Independent guide. Product names, features and data terms change often; check each vendor's current documentation before you rely on a feature. Nothing here is legal, security or professional advice.