Local guides · Post 6
AI for a small business in West Wales: what it costs and what it's worth
The honest price list for a small business in West Wales: what the £499 diagnostic buys, how builds are priced, and how to tell whether the hours are worth it.
Ask what AI costs a small business and you'll usually get either a sales page or a shrug. Here's the actual shape of it. The work is not buying a tool. It's finding the hours your week leaks, then handing the repeating ones over, with a person approving what goes out. The cost follows that shape: a fixed fee to find out where the hours are, a quoted price to fix the ones worth fixing, and a monthly layer that keeps the whole thing accurate.
Tishka AI is run from Cardigan, working with businesses across Ceredigion and Pembrokeshire. There's no client list yet, so this post sticks to prices and sums rather than promises.
The maths (a sketch, not a client figure)
Start with what your week loses. Pick one repeating job: say, an hour of admin a day, five days a week.
- 5 hours a week × £25 an hour = £125 a week
- × 46 working weeks = £5,750 a year
Now the front door. The £499 diagnostic finds, ranks and prices the hours that are recoverable:
- If it finds 5 hours a week, the fee has paid for itself within weeks.
- If it finds nothing worth automating, the fee comes back in full.
Builds are quoted from your own numbers before anything starts, so the spend is decided by you, in pounds, before work begins. After that a retainer keeps the system accurate month to month.
Score it: the three checks
- Hours. Four or more hours a week of repeating work. Below that, the fee outweighs the saving and the honest answer is: don't spend it.
- Repeat. The job must be writable. If a new starter couldn't follow it from a page, no tool can run it either.
- Risk-if-wrong. Recoverable slips score high. Anything where a wrong send costs a client scores low, and that work stays human anyway.
Try one thing this week (free)
Run the 20-minute count: list the jobs that repeat every week, guess the minutes each one takes, and total it against an hourly rate. You don't need a tool for this, only a pen and one honest hour. If the total lands near a week's wages a month, the diagnostic will pay for the conversation. If it comes to twenty minutes a week, close the tab and keep the £499.
Where a person stays
Every automation here starts as a draft. AI prepares, a named person approves, nothing sends unchecked. Financial, compliance and legal work stays approval-only, always. That rule is why the maths holds up: the hours that leave the week are the typing and the chasing, not the judgement. Anyone selling you a system where nobody checks the output is selling speed with your reputation attached.
FAQ
It's the front door. The diagnostic is one fixed fee and either finds 5 hours or comes back in full. Builds are scoped from your numbers and quoted before anything starts, so nothing is charged that wasn't agreed first.
The fee comes back in full, no conversation needed. A diagnostic that finds nothing has still done its job: it tells you where the week really goes.
No. The work connects the tools already in use. No new logins for staff, no data migration weekend, nothing to relearn.
For a one-step job that never changes, do exactly that. Where a partner earns the fee is wiring several tools together, keeping the knowledge accurate, and the person who maintains it after launch. DIY is the right answer for some jobs. This is the honest line, even though it costs us work.
It's built for businesses past the solo stage, roughly 2 to 50 people, where the owner still carries the company in their head. Location barely matters to the work itself, but being run from Cardigan means the same timezone, the same reference points, and a founder you can meet.