⏲️ The 20-Minute Job Safety Check

The worry about AI and your business expands to fill whatever you give it. Give it a sleepless night and it takes the whole night. Give it a scroll through the news and it ruins the scroll.

It never finishes, because it has nowhere to land.

So give it somewhere: twenty minutes, a timer, four questions. At the end you have a page telling you which parts of what you sell are still worth paying for in two years, which are quietly getting cheaper, and the one thing to change now.

Think of it like an MOT. You don’t MOT a car by lying awake wondering about the brakes. You put it on the ramp, run the checks, and get a sheet back.

The four questions

Five minutes each, in order, because each one feeds the next.

1. What did clients actually pay me for? Not your services page. Your last three months of invoices, line by line. Include the parts you never itemise: the chasing, the tweaks, the “quick call”, the reformatting, the thing you throw in free.

2. Which of those is getting cheaper? Some of what you sell, a client can now do a passable version of themselves in ten minutes. Some gets faster for you to deliver, which is good news until someone undercuts you with it. Some is untouched. You need that sorted honestly, and you are the worst person to sort it, which is why question two is the one you hand over.

3. What is my moat? The part they would still need you for even if they had the best tools in the world on their laptop. Relationships. Being accountable when it goes wrong. Taste. Knowing which of their ideas is the bad one. Showing up in person. Test each one with a hard question: does this need a human, or does it currently just involve one?

4. What do I change in the next fortnight? One move, with a day attached. Raise, repackage, drop, or add.

The whole thing as one prompt

Nick this one, fill in the blanks:

You're running a 20 minute business check with me. Be a blunt but kind mechanic: your job is an honest read on what I sell, not reassurance. Run these four stages IN ORDER, one at a time, waiting for my answer before you move on.

Me: I'm a [WHAT YOU DO] working with [WHO, e.g. "small builders and two estate agents"], mostly [HOW YOU CHARGE, e.g. "day rate, some monthly retainers"]. Been at it [YEARS].

STAGE 1, THE INVENTORY: ask me to list what clients actually paid me for over the last three months, taken from real invoices, not from my website. If my list looks tidy or too grand, prod me for the unglamorous bits I don't itemise: chasing, tweaks, quick calls, reformatting, the stuff I throw in free.

STAGE 2, THE EXPOSURE: sort each line into (a) a competent client could now do a passable version of this themselves, (b) it stays my job but it's getting faster and cheaper to deliver, so the price comes under pressure, (c) untouched. Be blunt. Don't flatter anything into group (c). Then tell me roughly what share of my income sits in (a) and (b).

STAGE 3, THE MOAT: ask what clients would still need me for if they had the best tools going. Challenge every answer with "does this NEED a human, or does it currently just involve one?" Get me down to two or three genuine ones, and say plainly if something I've claimed isn't a moat.

STAGE 4, THE VERDICT: give me a result sheet.
1. SOLID: what's genuinely safe, and why.
2. WATCH: what's exposed but not urgent.
3. CHANGE NOW: the single most important thing to act on.
4. ONE MOVE: one specific change for the next 14 days, with a suggested day, framed as raise, repackage, drop or add.
Then give me the result sheet as a document I can download.

Rules: British English, under 700 words, no pep talk. Judge only what I've told you, don't invent clients or figures, and don't predict the future with confidence. Where you're guessing, say "this is a guess". Ask me up to three questions before you start, and keep the whole thing moving.

Stage 1 decides how good the rest is. Weak: “design work, some consulting”. Strong: “three logo jobs, one full brand for a cafe, six sets of social templates, a website I still maintain, and about four hours a month of ‘can you just change this’ that I’ve never charged for”. From the first, it can only shrug. From the second, it can tell you something real.

Then put it in the diary

Six months from now, this will have moved. Two lines, once:

Nick this one, straight after:

You're writing me a prompt to use later. Give me a short version of that check I can save and re-run in six months, written so it reminds you what my verdict was this time and asks what actually changed. Under 150 words, ready to paste, nothing else in your reply. Ask me anything you need to know before you write it.

What twenty minutes can’t tell you

It is a snapshot, not a prophecy. It cannot see your local market, your reputation, or the client who would never leave you. It will be confidently wrong occasionally about what AI can and cannot do, in both directions, which is a decent reminder to put your own judgement on top of everything it says.

And it only works with the timer running. Read this and nod, and nothing changes.

Tonight

Do stage 1 on your own, in a note, before you open a chat. Fifteen lines from your last three months of invoices.

Most of the calm arrives right there, when the vague cloud of “my business” turns into a list you can point at.

P.s. if the verdict comes back “learn the tools properly”, The Claude Kickstart is the fastest honest way in: five real jobs off your plate and a playbook for your trade. £47, one weekend.