The 30-minute AI experiment to run before you leave the office on Friday

Here is a better use of Friday afternoon than reading about AI: actually trying it. Not buying anything, not committing to anything, not signing up for a course. Just a 30-minute experiment that tells you something real about whether AI is worth your attention.

Most business owners who feel stuck on AI have never done this. They have read about it, watched a video, perhaps had a brief play. But they have not taken a task from their actual working week and run it through a tool properly. That gap - between curious and tested - is where most of the "I'm not sure it's for me" feeling lives.

This experiment closes it.

What you need

A free ChatGPT account (chat.openai.com - free to sign up, no card required). A real task from your week. Thirty minutes.

That is it. No paid subscriptions. No integrations. No technical setup.

The experiment

Step 1: Pick a task (5 minutes)

Think about this week. What written task did you do that felt like it took longer than it should? It might be an email, a response to an enquiry, a piece of information you pulled together for a client, a social media post, meeting notes, a proposal, a policy document for a new team member.

The best task for this experiment has three qualities: it involves writing, it happens regularly (not once a year), and it currently takes you 15 minutes or more.

Examples that work well:

• A follow-up email after a client meeting
• A response to a new enquiry explaining what you do and how it works
• A job post for a role you are trying to fill
• A brief for a contractor or supplier
• A summary of a project or call for someone who was not there
• A social media post about something you know a lot about

Choose one. Write it down as a single sentence: "The task is: [what it is]."

Step 2: Describe the task properly (5 minutes)

Before you open ChatGPT, write down the inputs and the output. What information goes into this task? What does the finished thing look like?

For example: "The task is a post-meeting follow-up email to a new client. The inputs are: the client's name (Sarah), what we discussed (her website project, timeline of 8 weeks, budget of £4,500, next step is me sending her a detailed proposal by Wednesday), and the tone I want (professional but warm, not formal). The output is a 150-word email that confirms what we discussed and sets expectations for the next step."

This description is your prompt. The more specific it is, the better the result.

Step 3: Run it (10 minutes)

Open ChatGPT. Paste your description in. Read what it produces.

The first output will probably be 80-90% good. There will usually be a sentence or phrase that is not quite your voice, or a detail that is slightly off. This is normal and expected. AI tools are not trying to be you - they are giving you a strong first draft.

Now tell it what to change. "Make the tone a bit less formal." "The last paragraph is too long, can you shorten it?" "Add a line about the project deadline." You are having a conversation, not pressing a button once and hoping.

After one or two rounds of feedback, you should have something close to what you would have written. Note the total time it took.

Step 4: Score it honestly (10 minutes)

Ask yourself three questions.

Would I send this? If the answer is yes (perhaps with minor tweaks), that is a meaningful result. You produced something sendable in a fraction of the normal time.

How long would it have taken me without the tool? Compare that against what you just spent. The difference is your time saving per instance of this task. Multiply it by how often you do this task each week - that is the weekly saving from automating just this one thing.

Are there three other tasks in my week this might work for? If you can name three, you have just identified a realistic path to saving four or five hours a week. That is worth pursuing properly.

What the result means

If the experiment worked - if you got a usable output and could see how this would save you time every week - then AI is not just hype for your business. It is practical, it is accessible, and it starts with one task, not a company-wide rollout.

If the experiment felt clunky, or the output missed the mark in ways you could not correct with a simple instruction, the most likely reason is the task you chose. Some tasks require context that only you have - deep knowledge of a specific client relationship, a nuanced judgement call, a tone that is very specific to you. Try a different task: something more structural, more templatable, more generic. The experiment is worth another go.

If you are now curious about what a systematic look at your whole working week might surface - not just one task but all of them, prioritised by impact, with specific tools and setup instructions - that is exactly what the HoursBack AI Readiness Quiz gets you started on.

The one action for today

It is Friday. You have 30 minutes before the end of the day. Do the experiment. You will know more about whether AI is actually relevant to your business in those 30 minutes than in any amount of reading.

And if it clicks - if you finish the 30 minutes thinking "I could save four hours a week doing this properly" - sign up for the newsletter below and we will send you a practical guide on how to take that one working experiment and turn it into a genuine system.

Already done the experiment and want to know what to do next?

Take the free AI Readiness Quiz - it turns your experiment into a full action plan in 2 minutes.

Or browse the blog for more practical AI guides for UK business owners.