Run your own AI audit, or buy one? Here is how to do it yourself first
Short answer: you can absolutely do this yourself, and if you have a free weekend and some discipline you should try. What you buy when you pay for one is not insight you could never reach — it is the finishing of a job most owners start and abandon.
The DIY method, in full
No gatekeeping. This is genuinely the method.
1. Log a week. Five working days, a note on your phone, anything repetitive over ten minutes. Round to the half hour. Include the jobs you do at 9pm.
2. Cost an hour of your time. One number, conservative, written down. See how to work that out — this is the step people skip and it is the one that makes everything else decidable.
3. Group the log into jobs. You will end up with eight to fifteen named jobs. Put weekly hours against each.
4. Score each job for fit. Is the input mostly language? Is a checked draft useful? Those two questions sort the list; what AI is and is not good at goes through it properly.
5. Score each job for effort. Can you do it with a tool you already pay for and a written prompt, or does it need systems connected? Be pessimistic.
6. Plot the two scores. High value, low effort goes first. You now have an impact–effort matrix, which is all that phrase means.
7. Price the top six. For each: which tool, what it costs monthly, how long to set up, hours saved per week. Check current pricing on the vendor’s own site — do not trust a figure in any article, including this one.
8. Do the sum. Hours saved × your hourly rate × 46 working weeks, minus annual tool cost. Round down. If it does not clear your effort by a decent margin, it is not a quick win.
9. Write a four-day plan. Four tasks, each under two hours, one per day. Diary them.
10. Review in a month. Which stuck, which did not, and why.
What that costs you
Realistically: a week of background noticing, then four to six focused hours. At £55 an hour that is £220 to £330 of your own time, which is the honest comparison to make rather than pretending DIY is free.
Where DIY usually goes wrong
Three failure modes, and they are consistent.
It does not get finished. Steps one to three are easy and interesting. Steps seven and eight are tedious. Most owners have a half-filled list in Notes and no plan.
The savings get optimistic. Without a cap and a rule to round down, estimates drift upwards, and a plan built on cheerful numbers collapses at the first review.
You only consider tools you have heard of. Your shortlist is bounded by your own reading, which usually means the two or three brands with the biggest marketing budgets rather than the cheapest thing that does the job.
What buying one actually saves
Not magic. Three concrete things: the tedious half gets done rather than abandoned; the maths is capped, rounded down and printed so you can challenge it; and the tool shortlist comes from a fixed list of options that work for UK small businesses rather than from whatever you last read about.
That is what our AI Tools Report is — ten minutes of questions instead of six hours of spreadsheet, then a nine-slide costed report in your inbox within minutes, for £59.99 inc VAT. It is built on the playbook Green & Home Ltd uses in its own business, which is also why it is unglamorous about what is achievable.
Who should not buy it
Plainly, because we would rather you did not waste the money.
Do not buy it if you already have a clear plan. If you can name the three jobs, the tools and roughly what they cost, you have done this. Go and do the four-day plan instead.
Do not buy it if you have almost no admin. A sole trader taking jobs by phone, invoicing from a template, with no paperwork backlog and no quote pile, has very little for this to work on. The report will come back thin and you will feel short-changed — fairly.
Do not buy it if you enjoy the analysis. Some owners like this kind of thinking and are good at it. The method above is complete. Use it.
Do not buy it if your real problem is demand. If the phone is not ringing, saving admin hours is not your bottleneck.
For everyone else — plenty of admin, no written plan, and ten minutes easier to find than six hours — paying for the finished version is a reasonable trade. And if you want a person to argue with afterwards, the £299.99 report-and-call option adds 60 minutes and a refined report, credited against a build if you go further.
Questions, or want a tool we don't have yet? Email hello@grafto.co.uk — a real person replies.