How to Get Your First AI Win in 5 Minutes (Using the ICE Method)

The hardest AI minute is the first one. Not the tenth, not the hour-long fiddle with a prompt you nearly nailed — the first. Opening the tab. Typing something real. Most people I speak to already have an account with ChatGPT, Claude, or Gemini — and quietly, they have barely used it. The account is sitting there, like a gym membership you meant to start in January.

If that sounds familiar, you are in exactly the right place. The first time I sat down to actually use ChatGPT for something I would normally do myself, I spent 20 minutes staring at an empty prompt box, convinced I needed the perfect prompt before I could press enter. I did not. Nobody does.

This piece is short on purpose. By the time you finish reading, you will have a five-minute drill you can run today and walk away with a genuine win. Not a demo. A small, useful thing you actually needed done.

Before We Start: The Only Rule That Matters

Pick a task you already know the answer to.

That is it. That is the whole rule for your first attempt. If this feels anticlimactic, good — most first-try advice is bloated with prep work that exists to make the author feel thorough, not to make you successful.

Here is why it matters. When you pick a task you already know how to do — a weekly update you have written fifty times, a document you have skimmed before, a reply you could draft in your sleep — you can instantly tell whether the AI output is any good. You are not trying to evaluate unfamiliar work through a new tool at the same time. You are applying an expert eye (yours) to a familiar output (theirs). That is the fastest possible learning loop.

If you want help finding the right kind of task to give AI at work more broadly, our companion piece on finding your first real AI win at work walks through a three-question framework for choosing well. For today, though, do not optimise. Just pick something small and familiar. The perfect first task does not exist; the finished one does.

In our experience, people who run the 5-minute drill on a task they already know usually get a usable draft on the first try. People who pick an ambitious, unfamiliar task almost always abandon the session before minute three.

The 5-Minute Drill

Here is the whole thing, start to finish. Set a timer if it helps — it honestly does.

Step 1 — Pick a low-stakes task you already know (30 seconds)

Scan your current to-do list or inbox for something that fits three tests: you know how to do it, no one will die if the draft is slightly off, and it would take you more than 15 minutes to do from scratch. A weekly team update. A reply to a client asking the same question you have answered twice this month. A summary of a long document you half-read yesterday.

If nothing jumps out, use this default: draft a short message to a colleague explaining a meeting you just had. Everyone has one of those.

Step 2 — Open your tool (15 seconds)

ChatGPT, Claude, or Gemini — pick whichever you already have logged in. Do not switch tools, do not research "which is best," do not compare. The free tier is fine. The most-used AI tool is the one that is already open.

Step 3 — Write a 3-part brief (2 minutes)

This is the only part of the drill that actually requires thinking. Your brief needs three things, and only three:

1. What you want — the specific thing you need produced. ("A 150-word update for my team.")
2. Who it is for — the reader and their context. ("My team of six designers who know the project already but missed this week's standup.")
3. What format and tone — how it should look and sound. ("Short bullets under a one-line summary. Warm but professional — the way I actually talk, not corporate-speak.")

Paste your raw material at the end if you have any — notes, bullet points, the document you are summarising, the email you are replying to. Do not tidy it up first. Messy input is fine.

That is the brief. No clever framework, no secret keywords. If you want a deeper look at why this structure works and how to push it further, our guide on how to write AI prompts that actually work expands the idea. For your first win, three parts are enough.

Step 4 — Read it critically, then ask for one improvement (90 seconds)

Read what comes back like you would read a draft from a junior colleague. Two questions only:

• What is already good enough to keep?
• What is the single biggest problem with this version?

Then ask for exactly that one fix. "Make the tone more casual." "The second bullet is too vague — add a specific date." "Drop the opening greeting." One change, not five. Most first drafts do not need a full rewrite; they need one targeted nudge.

Step 5 — Ship it (60 seconds)

Copy the output. Paste it where it needs to go. Send, save, or tick the task off. Done.

You have used AI for a real piece of work. That is the win. It did not need to be transformational. It needed to happen.

🧠 Quick Challenge: You have five minutes and want a first AI win before your next meeting. Which of these is the best task to try?

• A) Draft the agenda for tomorrow's quarterly strategy review with the leadership team
• B) Summarise a 40-page industry report you have not read yet
• C) Rewrite a short reply to a client asking about lead times — a question you have answered three times this month

Three Tasks That Work Every Time for a First Try

If you want a shortlist rather than a to-do list scan, these three work for almost everyone. Pick the one closest to your actual week.

The weekly email or status update

You send a Friday update. You have sent dozens. Paste last week's version as a reference, paste this week's bullet points as raw material, and ask for a first draft in the same tone. Read it, nudge it once ("make it warmer" / "shorten the middle section"), send it.

Most first attempts at this save about 15 minutes and produce something structurally tighter than the hand-written version — because the AI is not tired on a Friday afternoon.

The long-document skim

You have a meeting Monday and a 20-page document you were meant to read over the weekend. Paste the document in (check it is not confidential first — see AI mistakes everyone makes if you are not sure what counts), and ask: "Summarise this in 200 words. Then give me the five questions a senior leader would ask about it."

You are not outsourcing the reading — you are getting a scaffold so your actual reading is faster and more targeted. Five minutes now saves roughly half an hour later.

The reword

Something you wrote sounds wrong. Too stiff, too casual, too long, not quite clear. Paste it in. "Rewrite this in a warmer tone, keeping all the facts. Under 100 words." Read the result. Keep 80%, fix the 20% that feels off.

This is the lowest-risk first win in the entire list. You already have the content. You are just using AI as a second pair of eyes. If you want ready-made prompts for this and similar tasks, the Prompt Library has templates you can copy straight into your tool of choice — skip the blank-page problem entirely.

Why This Works (The Improve Pillar in One Page)

The 5-minute drill is deliberately narrow. It is not the whole picture of what AI can do — it is one specific slice: using AI to improve work you are already doing. That slice is the Improve pillar of the ICE Method, and it is where almost every sustainable AI practice starts.

Improve has three properties that make it the ideal first win:

• Low cognitive load. You already know the task, so you can judge the output instantly.
• Fast feedback. A five-minute drill either worked or it did not. You learn in one sitting what would otherwise take a week of reading articles.
• Compounding return. A prompt that shapes this week's status update will shape next week's, and the one after. The second use takes 90 seconds. The tenth takes 30.

The Create and Educate pillars are real and worth your time — but they are second-win territory, not first. Create asks you to produce something you could not easily make on your own (a concept map, a first pass of a design, a voice clone of yourself). Educate uses AI as a tutor on something genuinely new. Both are brilliant. Both are harder for a first attempt because you cannot immediately tell if the output is any good.

Start with Improve. Build the muscle. The other two pillars become obvious once the first one clicks.

What to Do Tomorrow

You have your first win. Do not do anything dramatic with it. Do not buy a paid plan, do not read ten more tutorials, do not sign up for a course. One small next step is plenty.

Tomorrow, run the same drill on a different task. That is it. The goal of week one is not variety — it is repetition. Run the drill three or four times on three or four familiar tasks. Notice what consistently works for you and what does not. Save the briefs that produced the best outputs; they are your starter library.

If you want a more guided next step once you have three or four reps under your belt, our Learning Paths take the habit you have just started and extend it into a practical, sustainable practice. The Core Mindset path is the natural follow-up — it takes roughly an hour and removes most of the remaining "is this a good use of AI?" decision-making.

And if your second attempt goes sideways — if the output is rubbish, or you spend longer debugging than doing — that is not a failure, it is data. Most people hit two or three duds before a reliable pattern emerges. The people who get good at AI are the ones who shipped their first win before they talked themselves out of it.

Your First Win Is Five Minutes Away

Everything in this article takes about as long to do as it took to read. Pick one small, familiar task. Open the tool you already have. Write a 3-part brief. Read the output like a draft, nudge it once, ship it.

That is the whole drill. You have everything you need. The only thing standing between you and your first real AI win is the first minute — the one where you decide to start.

Ready to take your first win further? Our Learning Paths build on the 5-minute drill with step-by-step practice for real workplace tasks. Or browse the Prompt Library for ready-made prompts you can paste straight in. Both are free. Both are waiting. Start a Learning Path