AI can type out software faster than anyone. So why does a custom system still cost what it does? The honest answer, using a booking platform we built for a personal trainer.
You are not paying for the typing. AI does the typing now, and it is fast. You are paying for the things that were never in the brief, a design that does not look like everyone else's, testing it as every person who will use it, and somebody at the end of the phone when it needs to change. Here is what that looked like on a real job.
The job
A personal trainer with a busy diary. Some clients on packages, some pay as you go, some monthly. His coaching was already sorted on another app: programmes, check-ins, nutrition. What was missing was the business side. The bookings, the money, the paperwork.
He was planning to rent a booking system and make it work. Instead he wrote down everything he wanted and how he wanted it, and we built it. Online booking, card payments, health forms, reminders and a login for his clients.
It sits next to his coaching app rather than replacing it, because nobody wants to move all their clients onto a new app. It just takes the admin away: who is booked, who has paid, who has sessions left, and who still has not filled in their health questionnaire.
The ten things that were not in the brief
His brief was good. We still went through it line by line before anything was built, and changed over ten things. A few of them:
- Double bookings were not mentioned once. That is the first thing that goes wrong on any booking system.
- Health questionnaire answers are health data, and under UK data protection law health data gets extra protection. That decides how and where it is stored.
- The money goes from the client straight to the trainer, so every trainer needs their own Stripe account set up properly before their page can take a payment.
- No workouts and no nutrition. They already live in his coaching app, and keeping them out kept the system simple.
Knowing what to build, what to leave out and what was going to break is the part AI does not do for you.
Taste does not fit in a prompt
We built a concept first with a made-up trainer and made-up clients, so no real person's figures ever sat under a real name. The first version worked, but it looked plain, like every other booking app.
So it was given a direction: soft colours, frosted glass and one dark card that catches your eye. That became the look, and it has been the look ever since. You cannot describe taste in a prompt. You have to show it.
The other calls were ours too. The AI suggested building on an outside database service. We chose our own UK hosting, the same place we look after everything else. And our developer got a written handover, where it is at and what to watch for, so nothing lives in one person's head.
What it does
- The trainer drags sessions round his week, moves them or cancels them, and it tells him what happens to the money first. A credit goes back on the client's balance. A card payment does not refund itself, so it tells him what to do in Stripe. No surprises.
- He chooses whether the client hears about it by email, text or both.
- Every client has a card: sessions left, credits on or off, and a button to resend their password if they are locked out.
- Every service has its own rules: minimum notice, how far ahead people can book, and a form that has to be filled in before it is confirmed.
- Clients only see the hours he has open. Pick a session, pay, done. It lands in his Google Calendar, makes a Google Meet link for online sessions and sends a reminder before it starts.
- Clients get their own login to move or cancel within his rules.
Keeping AI honest
AI will tell you it has finished. It will tell you it has tested it. That does not mean it works. It can build a setting that saves perfectly and does nothing at all, a box you can tick with nothing wired up behind it. Looking at the screen, you would never know.
So we do not just look at it, we use it, as the trainer and as a client. And we know where the AI's checks stop. It cannot type passwords into websites, so it tests the demo account, and anything behind the trainer's real login gets checked by a person.
Then the real feedback starts
Going live is where rented software cannot keep up. Over one weekend the trainer sent through five changes, including cancelling a session from his side, taking credits off a client, stopping bookings with under 24 hours' notice and moving a session more than a week ahead. All five were live and tested on the Monday.
So what are you paying for?
- The ten things that were not in the brief.
- A design that does not look like everybody else's.
- Testing it as every person who will use it.
- Someone at the end of WhatsApp when it needs a change.
AI makes it affordable. Experience is still what makes it work.
Questions people ask
Is a custom booking system worth it over renting one?
If a rented one fits how you work, keep it. It is worth building your own when you are bending your business around the software, or juggling ways of paying it was never built for.
Can it work alongside the software I already use?
Yes, and often it should. This one sits next to the trainer's coaching app and leaves the coaching alone. The best system is the one that fills the gap, not the one that replaces everything.
