💷 What To Pay For And What To Skip With AI

The most profitable person in AI right now is a nervous beginner. By the time they’ve written their first proper prompt, someone’s sold them two chatbot subscriptions, a “1,000 prompts” PDF, and a £497 course with a countdown timer on the checkout page.

I teach AI for a living, and my most common piece of money advice is stop spending. Not because paying is always wrong - there’s a real moment where upgrading makes sense, and we’ll find yours - but because the order is backwards. The industry wants you to pay your way to competence. It doesn’t work like that. You use your way to competence, at £0, and pay only when the free plan starts genuinely costing you something.

So: an honest UK money guide. What free actually covers, the real signs you’ve outgrown it, and the three things I’d never spend a pound on. I run it all through the Three Wallet Rules.

Rule 1: Stay at £0 until it hurts

What it is: the free plan as your default state - not a trial, not a teaser, your actual plan until it visibly gets in your way.

How to do it: sign up at claude.ai and just… use it. Everything that makes AI feel personal is free: the profile where you introduce yourself, memory (yes, really - free for everyone), Styles, web search, the lot. Every getting-started guide on this site works entirely on the free plan. The main thing you’re “missing” is volume: free comes with usage limits, and when you hit one, Claude asks you to come back later - the limit resets after a few hours. That’s the trade. Full quality, capped quantity.

The bit nobody tells you: hitting the limit is information, not punishment. Most weeks, most beginners never touch it - which tells you the free plan fits. Hitting it repeatedly during real work tells you something too, and that’s Rule 1 doing its job: the free plan hurting is the only trustworthy upgrade signal there is. Everything else - the ads, the FOMO, the “serious people pay” energy - is noise.

Rule 2: One subscription, never a stack

What it is: when the day comes, you upgrade exactly one tool - the one you actually use - and you do not collect AI subscriptions like streaming services.

How to do it: paid plans for the big assistants all hover around £15-20 a month (prices shift, check the site before you believe me or anyone). One is a reasonable cost for a tool you use daily - mine’s Claude and it’s open from first coffee to log-off. Two is roughly £430 a year, and here’s the thing: they do the same job. Two chatbot subscriptions is paying twice for one assistant, like subscribing to two milkmen out of politeness.

The bit nobody tells you: the stack rarely comes from greed - it comes from headlines. Every month some model is newly “the best” and the twitchy move is to subscribe to the winner while keeping the old one “just in case”. Resist. The differences at the top are small and keep swapping, and your settled setup - the tool that knows you, with your memory and your instructions in it - beats a marginally cleverer stranger. If you genuinely switch allegiance, fine: cancel the old one first, then subscribe. Swap, never stack.

Rule 3: Never pay for what’s free inside the tool

What it is: the rule that saves beginners the most money. A whole industry sells you things the tool gives away, and knowing the pattern makes you unmuggable. The three I’d never buy:

Prompt packs. Those “1,000 ChatGPT prompts” PDFs and Etsy bundles. Ask the tool instead - “write me a great prompt for planning a hen do on a budget” - and it’ll produce a better one than the pack, tailored to you, for nothing. Buying prompts is buying bottled water while standing in a stream.

Shiny course funnels. The £297-£997 kind, with the countdown timer, the “founding member price ends tonight”, the income screenshots. What’s inside is, almost always, the content of free guides like this one, read aloud over slides. (Not all paid learning is a con - I write curriculum for publicly funded AI courses, so I’d hardly claim teaching has no value. But good teaching doesn’t need a fake ticking clock. The timer is the tell.)

“Access” resellers. Apps and sites charging monthly fees to use AI models you could use directly for free or cheaper. If an app’s main feature is being a wrapper around someone else’s AI with a mark-up, close the tab.

The bit nobody tells you: these all monetise the same feeling - “everyone else knows something I don’t”. They don’t. The people getting real value from AI are mostly using one assistant, on a normal plan, with good habits. The secret knowledge you’re worried about missing genuinely does not exist, and nobody flogging it through a countdown timer has it.

The five-question upgrade check

When you suspect you’ve outgrown free, don’t decide with your gut mid-annoyance. Answer these honestly:

Nick this one - the “should I upgrade?” checklist:

Yes or no:

1. Have I hit the free limit three or more times in the past fortnight?
2. Was I doing real work when it happened (a task with a deadline or a person waiting on it) rather than just having a play?
3. Did hitting the limit actually cost me something - a task abandoned, an evening rearranged, work done worse by hand?
4. Am I using AI most days, not just most weeks?
5. Have I already been using it for a month or more? (If it's week one, wait - early enthusiasm fades and limits feel bigger than they are.)

Score:
0-2 yes: stay free. You're not missing anything except the invoice.
3-4 yes: you're close. Give it two more weeks and count the limit-hits.
5 yes: upgrade the ONE tool you actually use, and cancel anything else AI-related you're paying for on the same day. Net cost: barely more than you were wasting.

Worked example, so you can see the bar: say you do the books for four local tradespeople. Every month-end you’re summarising receipts, chasing invoices and drafting reminder emails through Claude, and this month you hit the limit twice mid-flow on deadline day and finished at 11pm by hand. That’s questions 1 through 5 answered yes, and roughly £18 a month buying back several hours of month-end - an easy, calm yes. Compare: you hit the limit once, on a Sunday, while making a quiz for the pub. Stay free, enjoy the quiz.

Straight talk time

• I can’t see your usage, and prices move. Treat every number here as “roughly right in 2026” and check current pricing before deciding anything.
• Upgrading won’t make you better at AI. It removes the cap; it adds no skill. A paid plan used timidly is worth less than a free plan used well, and no subscription fixes vague asking.
• This is one honest opinion, not the market. Other tools’ free tiers differ, and if your job needs a specialist tool (design, video, code), that’s a separate question this guide isn’t answering.

If today only has five minutes in it

Audit before you upgrade. Tonight, find every AI-ish thing you’re already paying for - subscriptions you forgot, an app wrapper, that course you never finished - and cancel what you don’t use. Most people fund their eventual upgrade entirely from the clear-out, which means the answer to “should I pay for AI?” becomes: yes, £15-odd a month, and you’re still up on the deal.

Before paying for anything new, it’s worth seeing how far the genuinely free stuff stretches: Free AI Tools Worth Your Time. And if the money question behind your question is really “could AI help me earn?”, that’s the territory of Built For Freedom, my 30-day challenge (coming soon).