Anything with money in it gets a second look
Totals, VAT, discounts, payroll. Not because the studio is bad at arithmetic, but because a quiet error there compounds for months before anyone notices.
TUTORIALS / RUNNING IT FOR REAL
What it is bad at, and when to ask a person
Every page about a tool like this lists what it is good at. This one is the other half — the six things it is genuinely bad at, and the point where asking a person is not defeat but arithmetic.
5 minutes
Reading time
Five minutes now, instead of a bad week later
What you need
Wherever you are
Where you do it
01 Step by step
Read this before week three, not during it
None of these are secrets and none of them are fatal. They are the places where expectations and reality part company, and reading about them now is much cheaper than discovering them halfway through a build.
The pattern in all six is the same: the studio is excellent at writing software from a clear description, and no better than anyone else at the parts that are not software.
01
It can be confidently wrong
AI output is sometimes wrong in a way that reads perfectly well — the dangerous kind. Check what it built before you trust it with something that matters, and where a decision counts, have a person approve it rather than the machine decide alone.
02
It does not publish phone apps
What you get is web software that works properly in a phone browser. It does not put an app in the App Store or on Google Play. For most business systems this makes no practical difference; if it does for yours, know it now.
03
Messy migrations still need a person
Ten years of half-broken spreadsheets with three spellings of every client name will import badly on the first attempt, whoever does it. That is where our team earns its fee.
04
It is not a designer
What comes back is clean and consistent. If you need a brand-perfect public website with bespoke artwork, that is a different job and we will say so rather than take the money.
05
Some systems are barred outright
Medical diagnosis, legal opinions, hiring decisions taken without a person, biometric surveillance and more. The terms list them and the AI Act page explains where the line falls.
06
It cannot fix a process nobody understands
If the way the work flows is genuinely unclear inside the business, software will make that visible rather than solve it. Sometimes that is the useful outcome — but it is not the one people expect.
02 In your own words
When handing it over is the cheaper answer
Not everyone wants to sit down and describe their business to a screen, and there is nothing wrong with that. Our team will run the first build with you for a fixed fee — a conversation, a quote, a working system, your data in it, your staff trained — and then hand you the keys.
You end up in the same place either way: the same studio, the same system, changeable by you in plain language from that day on.
THE THREE SIGNS IT IS TIME
You have data in a state nobody can describe. You have tried the same correction three times and it keeps coming back wrong. Or the deadline is close enough that learning on the job is the expensive option.
Any one of those is a good reason to ask. None of them mean the platform was the wrong choice — the system you end up with is the same one either way.
03 What to watch for
A
B
A person approves, the machine proposes
Where the outcome affects a person — their money, their job, their health, their rights — build the system so a human makes the call. It is the right design and it is what the AI Act expects of you.
C
Ask before you spend a week
If you are unsure whether something is possible, ask. Describing and being asked questions is close to free; a week of building the wrong thing is not.
04 Where to go next
READ NEXT
Have it built for you — a fixed fee, a working system and the keys at the end.
The AI Act — which obligations are ours and which stay yours.