How to use ChatGPT or Claude in your daily work

Most people have tried ChatGPT or Claude at some point. Asked a question, got a reasonable answer, and then closed the tab again. The step that often gets missed comes next: giving it a fixed place in tasks you already do. That's where the time savings are.

Below you'll read how to go about it, without it getting technical. A few ways to start today.

Start small, with work you already do

The mistake most people make is seeing AI as a separate project. Something you have to free up time for, take a course in, make a plan for. So it stays at that one experiment.

It works better the other way around. Pick a task you do every week that costs you time. A weekly update you write. A long email thread you have to summarise. A meeting you have to prepare. Let AI help with it once and see what comes out. If you like it, you do it again the week after. That's how you build a habit, instead of a one-off trial that goes nowhere.

Compare it to a renovation. You don't start with the entire interior at once. You lay a foundation first, and here that foundation is that one recurring task.

Four kinds of work where AI already helps

In practice, most daily work falls into four kinds. AI helps with all four today.

Writing. A draft email, a job posting, a summary for a client. You provide the gist in a few sentences, AI makes a first version, and you edit. The blank screen is often the hardest part; AI takes that off your hands.

Summarising and analysing. A thirty-page report, the minutes of a meeting, a string of emails about one case. You let AI pull out the main points, and you ask follow-up questions on what isn't clear. You then read the heavy material with focus, instead of in full.

Preparing. A difficult conversation, a presentation, a tricky email. Ask AI which questions a client might raise, or which counterarguments you can expect. You stand firmer in the conversation because you've thought about the responses in advance.

Finding things out and learning. A topic that's new to you, explained in plain language, at your level. If you don't get something, you ask for it more simply. You don't need a colleague who happens to have time for it.

Giving a good instruction

The quality of the answer depends on the quality of your question. A vague request gives a vague result, just as a contractor told to "make something nice of it" can go in any direction.

A usable instruction usually contains four things:

• Who you are or who it's for. "I'm a recruiter" or "this is for a client in construction."

• What you want. "Write an invitation" or "summarise this."

• In what form. "In three short paragraphs" or "as a bulleted list" or "as a business email."

• What the tone should be. "Friendly but businesslike" or "short and direct."

An example. Instead of "write an email about our workshop", this works better: "Write a short, friendly invitation email to an entrepreneur I spoke with last week, for a workshop about AI in daily work. Five sentences at most, not salesy." The second gives you something you can almost send. The first gives a generic text you still have to rewrite completely.

You don't have to get this right in one go. If you don't like the answer, you just say what you want different. "Shorter", "less formal", "don't mention the amount." It's a conversation, not a form.

What to watch with your data

Two things deserve attention once you put AI to serious use for work.

The first is your data. The moment you put personal or confidential information into an AI tool, the GDPR and the agreements you have with clients apply. With free versions, your input can be used to improve the model. In business subscriptions you can often set it so this doesn't happen. Check that per tool before you put client data into it, and when in doubt, don't put names or case files into a free version.

The second is control. AI can say something firmly and convincingly that still isn't correct. A wrong figure, a source that doesn't exist, an assumption based on nothing. For a draft email that matters little, since you read it over anyway. For a quote or a legal text, you want to check everything before it goes out. You remain responsible for what you send, not the tool.

ChatGPT, Claude or something else

A question almost everyone asks: which one should I get? ChatGPT, Claude, Copilot, they resemble each other and for many daily tasks it makes little difference which you pick. For more specific work there are differences, for example in writing style, in working with long documents, or in the link to your own files.

If you want to choose calmly before you begin, compare the tools on those points. If you just want to start, take the tool you already have access to and begin with that one recurring task.

How we can help

Reading how it works is one thing. Applying it to your own work is another, and that's where people often get stuck. Which tasks do you hand to AI, how do you give a good instruction for your kind of work, and how do you make sure it doesn't fade again after a week.

For that we have the workshop Claude as Your Daily Work Partner. In one afternoon you work with your own documents and tasks, and you go home with a few methods you'll use the next day. No theory upfront, you do it straight away on your own work. The workshop can be followed via open enrolment or in-company, and is usually payable from your training budget.

Read more in this guide: ChatGPT, Claude or Copilot: which fits which work and how to give AI good context and goals.

Frequently asked questions