Muck Around in Private, Then Bring It Back to Work

It is a Saturday night. I am at a friend's place. The kids have handwritten their Maccas order on a scrap of paper, and the dad is hunched over the app, entering it one item at a time. A cheeseburger here. A frozen Coke there. Building the whole order by hand.

I take a photo of the handwritten list. I drop it into ChatGPT with one instruction. Read this, find the cheapest way to buy all of it, and bundle it into meal deals where that works out cheaper. Print the list back first so we can check the handwriting.

It transcribes the kids' scrawl. It prices everything line by line. Then it rebundles the singles into meal deals. Close to 20% cheaper, for the exact same food.

The dad looked at me like I had done a magic trick. I had not. I had done at McDonald's what I do at work every day.

This is the thing almost nobody says out loud. The people getting real value from AI at work are almost always the ones treating it like a toy at home. Personal play is not a waste of time. It is the on-ramp nobody talks about.

You cannot train your way to fluency

Most organisations think AI capability comes from a course. Book the workshop, tick the box, wait for productivity to go up.

The workshop matters. I run them for a living. But a workshop gives you the map. Fluency comes from the kilometres you put in afterwards, and most of those kilometres happen outside of work, on low-stakes problems where it does not matter if you get it wrong.

Think about how anyone got good at anything. Not from the lesson. From mucking around between the lessons. The kid who becomes a guitarist is the one playing in their bedroom for fun, not the one who only touches the instrument during the Tuesday class.

AI is the same. The people who only open it at work, under pressure, with something real on the line, stay shallow. They type a question, get a reply, and decide AI is a slightly fancier search engine. The people who play with it on the weekend build something the work crowd never does. Pattern recognition. An instinct for what these tools can and cannot do. The confidence to push past the first answer.

That instinct is the whole game. And you cannot download it. You have to earn it on problems you actually care about.

Three kinds of home play that build workplace muscle

Not all play is equal. Here are the three that transfer straight back into work.

Comparison play

Point AI at a decision you have been avoiding.

I had the same private health insurance since I got married. Never changed it. I loaded the policy and my full claims history into Claude Cowork and asked it to find me something better. What came back was not a suggestion. It was a built dashboard. Current premium, what I was using versus what I was paying for, shortlisted providers, estimated savings, and a call script for the awkward phone calls.

That is the exact muscle a procurement lead needs for a vendor review. Same move. Same prompt structure. The only difference is the data. Practise it on your health insurance and you have practised it on your supplier contracts.

Creative play

Make AI do something useless and fun. The skill it builds is anything but useless.

I once briefed Gemini's storybook tool to write a children's book about Charlie, a seven year old whose mum is a CFO and dad is a CPO. Pure silliness. The reward was learning, in my hands, how to brief for tone, character, and constraint, and how to push a tool past its first generic attempt.

A CFO who watched me do that went back to his office and ran his next board report through the same kind of prompt to find a fresh narrative arc through the data. He did not present a storybook to the board. He used the play to sharpen the real thing. The silly skill paid the serious bill.

Agent play

Hand AI a real task and let it act, not just answer.

I have four kids. Planning a holiday means finding six frequent flyer seats on the same flight, which is a nightmare by hand. So I logged into my airline account, gave an AI agent the brief, Melbourne to the Gold Coast, six seats, seven nights, flexible on dates, and let it work through the booking screens and come back with ranked options.

A low-stakes family trip taught me exactly how to scope a task for an agent, how much rope to give it, and where it needs a human to check the work. That is the same judgment you need before you point an agent at anything that matters in the business.

The objection, and why it is backwards

"I do not have time to play with AI."

I hear this from senior people constantly, and it has the logic exactly upside down. You do not have time precisely because you have not built the fluency that play creates. The fifteen minutes you spend optimising a Maccas order is not stolen from real work. It is the cheapest training you will ever do for the real work.

The people who feel they have no time for AI are the ones who will keep having no time. The ones who treat it as a toy on Saturday are the ones quietly getting their Tuesdays back.

Your one-week challenge

You do not need a strategy for this. You need a week.

Pick one low-stakes thing in your own life and hand it to AI properly. Not a one-line question. A real brief, with context and a clear outcome you want.

• Photograph something handwritten and have AI turn it into something useful
• Load a personal decision you have been putting off and ask for the full action plan, including the awkward steps
• Give an agent a small real-world errand and watch where it shines and where it stumbles

Do one a day for a week. By Friday you will not have read about AI capability. You will have built some. And the next time something real lands on your desk at work, you will already know exactly how to brief it.

The point underneath

We treat play as the opposite of work. With AI it is the on-ramp to it.

The fluency that closes the Ambition Gap at work does not get built in a boardroom. It gets built on a Saturday night, on a problem that does not matter, by someone who was curious enough to muck around. Give your people permission to play. The serious returns are hiding inside the silly experiments.