How to use AI at work: 15 small tasks beginners can try today

Most advice about using AI begins with the tool: choose the best model, learn the perfect prompt, master a list of features. I think that puts the cart before the horse.

When someone asks me how to use AI at work, I would rather begin with a different question: what small job keeps taking your time even though you already know how it should be done? It might be turning rough notes into an email, following up a quotation or making sense of a voice note sent from the field.

You do not need to understand computer programming to begin. If you can explain a job to a new colleague, read their first attempt and point out what needs correcting, you already understand the basic working relationship.

This guide gives you 15 small tasks you can try with Gemini, ChatGPT, Copilot or another AI assistant available to you. Please do not try all 15 today. That is how a productivity experiment becomes a second job. Choose the one boring or repeated task that would make today genuinely easier.

Quick answer for a busy reader

• Begin with one small, low-risk task.
• Use a tool you can already access before paying for another one.
• Give AI enough context to understand the job.
• Remove private or confidential information you do not need to share.
• Ask for a short, useful format.
• Treat every answer as a draft.
• Check facts, numbers, names, dates, links, prices, and promises.
• Measure whether the task became faster or better.

AI is most useful when it reduces the burden of preparation. You remain responsible for the final message, decision, or action.

What an AI assistant can and cannot do

An AI assistant works through ordinary written instructions. You can ask it to summarise, rewrite, organise, compare, explain, or prepare a draft.

It can produce a confident answer that is incomplete or wrong. It does not automatically know your company, customer, prices, current stock, policies, or local requirements. It only knows the information available to it, which may not include the most recent or correct facts.

Think of AI as a quick assistant that needs clear instructions and supervision. It can produce a page before you have finished making tea, but speed and wisdom are not the same thing. It is not the owner, accountant, manager, lawyer, doctor, or final decision-maker.

The mental model I find most useful is simple: you provide the judgement; AI helps with the preparation. Sometimes the preparation is most of the work. Sometimes it is the easy part. Knowing the difference is more important than knowing the name of every AI model.

Before you try the first task

Use information you are allowed to share

Do not paste passwords, payment credentials, identity documents, private employee information, confidential customer records, medical information, or restricted company documents into an AI tool.

Remove names and identifying details when they are not needed. Follow your employer's rules and the privacy settings of the tool you use.

Use a simple four-part request

You do not need a clever prompt, and you are not casting a spell. Tell AI:

1. The job: what you want help preparing.
2. The facts: what it should know.
3. The limits: what it must not guess or change.
4. The output: the format and length you want.

For example:

Help me write a short email asking a supplier to confirm a delivery date.

The order was placed on Monday and the expected delivery was Friday.

Do not sound angry and do not invent an order number or new deadline.

Write one email under 100 words with a clear subject line.

This pattern works for nearly every task in this guide.

Start with the work that already annoys you

The tasks below are not tricks for showing that AI can generate text. Each one begins with work people already do and gives you a source you can use to check the result. The first group is about organising your own work; then we move into customers and business; finally, we use AI to support your career.

1. Turn rough notes into a professional email

You may know what you need to say but struggle to arrange it clearly. Give AI your rough points and ask for a short email.

Try this:

Turn these notes into a clear professional email:

• thank the customer for the meeting;
• confirm that we will send the quotation tomorrow;
• ask them to confirm the delivery location.

Use warm, plain English. Do not add a price or make any new promise.

Check the recipient, dates, commitments, tone, and any facts AI added.

2. Make a long email easier to understand

Paste only the part you are allowed to share and ask for the main points, required actions, and unanswered questions.

Try this:

Explain this email in plain language.

Return:

1. What the sender wants
2. What I need to do
3. Important dates
4. Questions I should ask

Do not act on the summary without checking the original message, especially when money, contracts, deadlines, or customer commitments are involved.

3. Turn meeting notes into actions

Untidy notes often mix decisions, ideas, questions, and side conversations. AI can prepare a first action list.

This is particularly useful when the notes say “someone will follow up.” Someone is a remarkably busy employee in many organisations, but they rarely appear on the payroll.

Try this:

Turn these meeting notes into a table with:

• decision;
• action;
• owner;
• deadline; and
• open question.

Do not assign an owner or deadline unless it appears in the notes. Mark missing information as "not confirmed".

Check the result with someone who attended the meeting. For a full workflow, use the guide on turning meeting notes into a clear action list.

4. Plan your workday from a task list

AI can help group tasks, identify dependencies, and prepare a realistic order. It does not know every interruption or priority unless you explain them.

Try this:

Help me organise today's work.

I have six hours available. These tasks must be completed today: [task list]

Group similar tasks, put urgent customer work first, and leave one hour for unexpected work. Ask me about anything that is unclear.

Review the plan against real deadlines and your energy, travel, meetings, and responsibilities.

If the plan quietly assumes that your six-hour day contains eleven hours, ask it to try again. AI is very optimistic about other people's calendars.

The first four tasks are deliberately ordinary. That is a strength, not a lack of ambition. When you understand the work and can spot a poor answer, you can experiment without handing AI a decision you are not ready to supervise.

5. Prepare a meeting agenda

Give AI the meeting purpose, participants, available time, and decisions needed.

Try this:

Prepare a 30-minute meeting agenda.

Purpose: decide how we will reduce late customer deliveries. Participants: owner, sales lead, warehouse lead, and driver.

Include time for the current problem, causes, possible changes, decision, owners, and next review date. Keep the agenda practical.

Remove unnecessary items and send the agenda early enough for participants to prepare.

6. Turn a voice-note transcript into a useful update

WhatsApp voice notes often contain important details mixed with repetition. A “small update” can take seven minutes and reveal the actual action somewhere around minute five. If you have a transcript you are permitted to use, ask AI to organise it.

Try this:

Turn this voice-note transcript into:

• work completed;
• work still open;
• problems;
• people who need to respond; and
• next actions.

Do not invent names, dates, amounts, or commitments.

Listen to the original voice note when a word, number, location, or customer name is unclear.

7. Create a checklist from a task you know

An experienced employee may know a process that has never been written down. Ask AI to interview the person or organise their explanation.

Try this:

Turn these notes into a beginner-friendly checklist for opening our shop.

Keep the steps in the order provided. Separate safety checks, customer-area preparation, stock checks, and cash-opening steps. Mark anything that needs owner approval.

Ask the experienced employee to test the checklist in real work. AI can organise knowledge, but the people doing the job must confirm that it is correct.

8. Prepare a weekly report from updates

Give AI checked numbers and short notes, then ask it to separate results, problems, and next actions.

A spreadsheet can contain every number and still refuse to answer the manager's question. The useful job here is turning rows into a short account of what changed—not decorating the report with confident adjectives.

Try this:

Prepare a one-page weekly business update from the information below.

Use these headings:

• Results
• Customer issues
• Work not completed
• Decisions needed
• Priorities for next week

Use only the facts and numbers I provide. Flag conflicting figures.

Compare every number with the original source. The detailed weekly spreadsheet reporting guide shows how to review the result.

So far, we have used AI mostly to organise information you already have. The next group brings customers, suppliers and the public into the picture. The checking standard needs to rise because another person may act on what you send.

9. Draft a customer reply

AI can prepare a reply using approved facts. It should not invent prices, stock, discounts, delivery times, warranties, or company policies.

Try this:

Write a WhatsApp reply to a customer asking whether this product is available.

Checked facts:

• Product: [name]
• Availability: [status]
• Price: [approved price]
• Collection or delivery area: [area]

Use warm, plain English and end with one useful question. Do not add a discount or delivery promise.

Read the reply as the customer would. Correct anything that sounds generic, pushy, or unlike your business.

If the message sounds as though it was written for “Dear Valued Human,” rewrite it. Customers do not need to know that AI helped; they do need to feel that the business understood the question.

10. Follow up a quotation politely

Many people delay follow-up because they do not know what to say. AI can prepare several short options.

Try this:

Write three short WhatsApp follow-ups for a quotation sent three days ago.

The goal is to ask whether the customer has questions and whether the work is still being considered. Do not create urgency, offer a discount, or assume the customer has made a decision.

Choose one message and add any relevant context. Do not send a long sequence without considering the customer's response.

One respectful follow-up can help. Six versions of “just checking in” will not turn silence into a sales strategy.

11. Turn repeated questions into a customer FAQ

Collect five to ten common questions without unnecessary customer information. Add the business's checked answers. The same question often returns wearing different clothes; AI is useful for grouping the variations without pretending the answer is new.

Try this:

Organise these customer questions and approved answers into a simple FAQ.

Combine questions that mean the same thing. Use plain English. Do not change prices, policies, service areas, or product facts. List any question that does not have an approved answer.

Ask the person responsible for customers to review the FAQ. The customer-reply helper guide shows how to reuse approved answers.

12. Create marketing ideas from real business work

Instead of asking AI to invent advertising claims, give it a real product, service, installation, customer question, or useful lesson.

Try this:

Create five WhatsApp Status ideas from this completed job.

Facts: [checked description of the work]

Audience: [customer type]

Use one idea to explain the problem, one to show the process, one to answer a common question, one to show the result, and one to invite an enquiry.

Do not invent a testimonial, price, saving, or customer reaction.

Use only photos you have permission to publish. Check every public claim before posting.

AI was not at the installation, inside the shop, or on the call with the customer. If a claim is not in your source material, enthusiasm is not evidence.

13. Compare supplier quotations

AI can organise differences between quotations, but it should not choose a supplier without your judgement.

Try this:

Compare these supplier quotations in a table.

Include:

• item;
• quantity;
• unit price;
• total;
• delivery date;
• payment terms;
• warranty; and
• information that is missing.

Do not recommend a winner. Flag figures that do not add up.

Remove confidential details you do not need to share. Check all figures against the original quotations and consider quality, reliability, service, and risk—not price alone.

The lowest figure is easy to circle. Whether it becomes the most expensive sequel is the part that still needs business judgement.

14. Improve a CV without inventing experience

AI can make real experience clearer. It must not create qualifications, employers, dates, achievements, or skills you do not have.

Try this:

Improve this CV bullet using clear, specific language.

Original: "Helped customers and did stock work."

Ask me questions about what I actually did before rewriting it. Do not invent numbers, achievements, software, or responsibilities.

The strongest CV is accurate and easy to understand. Confirm that you can explain every sentence in an interview.

This matters because an interviewer is allowed to ask a follow-up question, an inconvenient feature that no prompt can disable.

The final two tasks are personal. AI can help you express your experience and practise using it, but it should not manufacture a more impressive version of you. The useful outcome is clarity and confidence—not a story you cannot defend when somebody asks a follow-up question.

15. Practise for a job interview

Give AI the job description and a version of your CV with unnecessary personal information removed.

Try this:

Act as an interview coach for this role.

Ask one question at a time. After each answer:

• tell me what was clear;
• show what evidence is missing;
• ask one follow-up question; and
• help me improve the answer without inventing experience.

Include questions about the work, teamwork, difficult situations, and why I want the role.

Say your answers aloud. A natural, truthful answer is more useful than memorising an AI-written speech.

You want to sound prepared, not as though somebody pressed play.

A seven-day beginner plan

You can build confidence without trying everything at once. There is no certificate for opening fifteen AI chats before lunch:

1. Day one: rewrite one ordinary email.
2. Day two: summarise one non-confidential document.
3. Day three: turn rough notes into an action list.
4. Day four: prepare one customer reply using checked facts.
5. Day five: create a checklist for a task you understand.
6. Day six: compare the AI draft with how you normally do the work.
7. Day seven: choose one task worth repeating and write down the checking steps.

Keep the original work and the AI-assisted version during the test. Record the time taken, corrections required, and whether the finished result became more useful.

How to check any AI-generated work

Before using the result, ask:

• Are the names, dates, amounts, and contact details correct?
• Did AI add a fact I did not provide?
• Does the message promise anything I cannot approve?
• Can I find the source behind every important claim?
• Does the tone suit the customer, colleague, or manager?
• Have I removed unnecessary private information?
• Would I be comfortable taking responsibility for this result?

If the answer to the last question is no, do not send or use it yet.

When AI is the wrong tool

Do not use a general AI assistant as the final decision-maker for hiring, firing, lending, medical care, legal rights, tax treatment, safety, or other high-impact decisions.

Do not use it to hide a lack of evidence, create a false customer testimonial, imitate someone's identity, or produce documents that pretend to be genuine records.

Sometimes the faster and safer action is to ask the responsible person, read the original document, call the customer, or use the approved company system.

Start with work you already understand

I do not think the goal is to use AI every day or to force it into every part of your work. The goal is to recognise the moments where it removes a burden without removing your judgement.

You will learn more from using AI on one familiar task than from reading fifty lists of advanced tools (including this one, if you never try anything). Choose a small job, give clear information, check the draft and notice what changed.

When the same workflow works repeatedly, you can decide whether to save the instructions, train a colleague, connect it to other tools, or explore a more focused AI assistant.

And if the old way remains faster or produces a better result? Keep the old way. A useful tool should earn its place in your work.

If you are choosing a first business workflow, continue with which AI agent should your business start with?.