Skip to content
Search
8 min read

How to Use AI at Work: Your First Week, Five Tasks to Try and Three to Keep

Meelie Gagne

TL;DR: The fastest way to learn how to use AI at work is to give it five low-risk jobs this week (rewriting, summarizing, outlining, preparing for a decision and drafting a checklist), keep three jobs for people (anything with customer personal data, final numbers and decisions about people), and check everything it gives you before anyone outside your team sees it.

If you own a business or run a team, you've probably been told that AI will change how your people work. That may be true, but it doesn't tell you what to do on Monday. This lesson is about how to use AI at work in the first week: what to hand over, what to keep, and how to tell the difference. It assumes you've never written a prompt. A prompt is just the message you type to the AI assistant. The assistant is a chat tool built on a large language model, which is software trained on a huge amount of text to predict useful replies.

Before you start, check one thing. If your company has an AI policy, follow it, including which assistant you're allowed to use. If it doesn't have one, the rules in this lesson are a sensible place to start, and the last section shows you how to turn them into a policy.

What people actually use AI for at work

You don't have to guess where AI helps most. Two large studies give a clear picture.

Researchers at OpenAI, Harvard and Duke classified a sample of messages that consumer-plan users sent to OpenAI's chat assistant and found that writing was the most common use at work, at 40% of work-related messages in June 2025. Across all writing requests in the sample, work and personal, about two thirds asked the assistant to edit, critique, summarize or translate text the person had already written, rather than write something new from nothing (Chatterji et al., "How People Use ChatGPT", OpenAI, September 2025). The same paper found that about 81% of work-related messages fell into two kinds of work: getting and interpreting information, and making decisions or solving problems.

Pew Research Center surveyed US workers and found the same pattern from the other side. Among workers who had used AI chatbots for their job, 57% used them for research or finding information, 52% for editing written content and 47% for drafting documents. Forty percent of those users said chatbots were extremely or very helpful for getting things done more quickly, and 29% said the same about improving the quality of their work (Pew Research Center, February 2025).

So the first week should start where the evidence is strongest: text your team already writes and information it already reads.

Five tasks to try this week

Each of these uses text you already have, has a person checking the result, and costs nothing if the first try is poor.

1. Rewrite something you already wrote

Take an email, a job ad or an announcement you've drafted and ask the assistant to make it clearer. This is the use the research shows most often, and it's the easiest to check because you know what you meant.

The Prompt:

Here is an email I wrote to my team about a schedule change.
Rewrite it so it is clear and friendly, under 150 words.
Keep every date and time exactly as I wrote them.
List anything you changed in meaning, if anything.

[paste your email]

Why This Works: You've given it a job, a length, a rule it must keep and a way to show its work. The last line makes it tell you if it changed something you didn't ask it to change.

2. Summarize a long document

Paste a supplier contract summary, an industry report or a long internal update and ask for the five points that matter to your business, each with the sentence it came from. Asking for the source sentence lets you check each point in seconds.

3. Outline a plan you've been putting off

Describe the project in a few sentences and ask for an outline with steps, owners and open questions. Expect to cut or change parts of it. What's left is a starting point you didn't have before.

4. Prepare for a decision

Ask it to list the questions you should answer before you decide something, such as switching a supplier or opening on Saturdays. You make the decision. The assistant only helps you see what you'd want to know first.

5. Turn a habit into a checklist

Describe how you open the shop, onboard a new hire or close the month, and ask for a numbered checklist. Then hand it to the person who does the job and ask what's missing. The operations track goes further with this.

Three jobs to keep for people

These aren't forbidden forever. They need rules, tools and checks that a first week doesn't have.

1. Anything with customer or employee personal data

Don't paste names with addresses, phone numbers, account numbers, health details, pay or anything else that identifies a person. The same goes for passwords, access keys and other secrets. Even an approved tool may keep what you paste, depending on its settings and your company's agreement with the vendor. Strip the details first, or don't use AI for that task yet.

2. Final numbers

AI assistants can write a figure that looks right and isn't. That's called a hallucination: the model produces a confident answer that isn't supported by any source. Let it explain a spreadsheet or draft the words around the numbers, but the totals in an invoice, a quote, payroll or a tax filing come from your systems and a person checks them.

3. Decisions about people

Hiring, firing, pay, discipline and performance reviews stay with a person. An assistant can help you write a clear job post or tidy notes you wrote yourself. It shouldn't rank candidates or judge someone's work for you.

How to tell the difference on any new task

When someone on your team asks "can I use AI for this?", three questions cover most cases.

  1. Is there personal data or a secret in it? If yes, remove it or don't use AI.
  2. Will a person check the result before it reaches a customer, a regulator or a bank? If no, don't use AI for it yet.
  3. If the assistant gets it wrong, can you catch it and fix it cheaply? If not, keep it for people.

A task that passes all three is a good candidate. This is the same habit a technical team builds into its systems, which the verification prompt pattern describes for developers.

A worked example: the weekly update

Here's a full first-week task from start to finish. You have notes from three team leads and need a short update for the whole company.

The Prompt:

You are helping a small business owner write a weekly update.
Below are notes from three team leads. Write an update for all
staff, under 250 words, with three short sections: what went
well, what's changing next week, and what we need help with.

Rules:
- Use only facts in the notes. Do not add numbers or names.
- If two notes disagree, point it out at the end instead of choosing.
- Plain, warm tone. No jargon.

Notes:
[paste the three sets of notes, with any customer names removed]

Why This Works: The first line gives the assistant a role. The rules stop it inventing figures, which is a common way a summary goes wrong, and the disagreement rule surfaces a problem you'd want to know about. Setting the length and the sections gives you something short enough to send as it is.

What you get back: A draft update in three sections, plus a short note if the leads' notes conflict. Read it once against the notes, fix anything that's off and send it.

Turn the week into a short policy

At the end of the week you'll know which tasks worked. Write them down in one page:

  • which assistant the team may use, and that it's the only one for work data
  • the three jobs that stay with people
  • what never gets pasted in: personal data, passwords, keys, contract terms you're bound to keep confidential
  • that a person checks every fact, figure and name before anything goes out
  • who to ask when a task isn't on the list

If you want a fuller framework later, the US National Institute of Standards and Technology publishes a voluntary AI Risk Management Framework and a profile for generative AI. Most small teams won't need it in week one, but it's a reliable reference when the list grows.

For more examples by job, see the other tracks in Start here, and the guides in the Applied pillar.

Frequently asked questions

How do I start using AI at work?

Pick one task you already do every week that's mostly text, such as rewriting an email or summarizing a report. Use the assistant your company allows, remove any personal data, and check the result before you use it. Add a second task once the first one saves you time.

What should you not use AI for at work?

Don't use it on customer or employee personal data, passwords or other secrets, on final numbers such as invoices or payroll, or to make decisions about people such as hiring or pay. Those need rules and checks that go beyond a chat window.

Is it safe to use AI for work emails?

It's generally fine for rewriting and tidying emails if you follow your company's AI policy and leave out personal details, account numbers and confidential terms. Read the draft before sending, because the assistant can change a date or a promise without meaning to.

Will AI replace my team?

The research above shows people mostly use AI to edit, summarize and look things up, and to get help thinking through decisions. Treat it as a way to give your team time back on routine writing, and keep the judgment with them.

Do I need to know how to code to use AI at work?

No. Everything in this lesson works by typing plain sentences into a chat window. The glossary explains the technical words you'll hear if you want to go further.

Rather have someone set it up for you?

That's a fair call. LTFI, the agentic business operating system from Kief Studio, can set up the work these guides describe and run it with you. No website yet? Kief Studio will build you one for free.