AI-CLOUD.IT

We work with leading AI partners.

Claude Gemini OpenAI AWS Vertex AI Databricks DeepSeek Cerebras Qwen Stability AI

Guide

Can ChatGPT build software for your company? Yes, but here is what is missing

Lots of people are asking: if ChatGPT writes code, why pay anyone for a business system? To be fair, it really does write the code, and the code often works. Between that code in a chat and a program your company relies on every day, though, sit hosting, data, logins, maintenance and security. This guide takes them one at a time, without blowing the problems up or playing them down.

The short answer: yes, although code on its own is not enough

A chat assistant such as ChatGPT, from OpenAI, gives you code in a conversation: you ask in words and it writes it. For a small script, a spreadsheet formula or a one-off calculation, that is genuinely enough, and it would be wrong to say otherwise. A business system is another matter, because it is a program that has to be running every morning, reachable at an address, with the company’s data in it, with colleagues signing in with their own password and seeing only what they should. It has to be backed up, updated and fixed when something changes. A chat’s reply contains none of that, and somebody will have to take care of it afterwards.

  • For a small job you do once, the chat will do.
  • For a program several people use every day with real data, you also need everything that comes after the code.

What ChatGPT does well

Before talking about what is missing, it is fair to say what is there. A chat assistant is a useful tool even for people who do not program, and many companies already use it well. Some versions can also run or preview small pieces of code, so you can quickly tell whether an idea stands up. For quite a few tasks, that is all you will ever need.

  • It explains a technical idea in plain words, patiently and without judging.
  • It writes a complicated spreadsheet formula, a macro or a small script to rename files or merge tables.
  • It suggests how to organise data: which tables, which fields, what links to what.
  • It drafts a specification: what the program must do, who uses it, which special cases exist.
  • It writes a first version of code to start from, if someone in the company can read it.
  • It helps write the texts, emails and descriptions that later end up inside the program.

Hosting: where the program runs, and who keeps it on

The code you get in a chat runs nowhere until someone puts it on a server. You have to choose where to host it, configure it, give it an address (perhaps the company’s own domain) and a certificate so the connection is encrypted (HTTPS). Then it has to stay on: if the server stops at night or during the holidays, nobody can work the next morning, and someone has to notice and restart it. An internal program the whole company uses cannot depend on the laptop of whoever built it. One more question to the chat will not help here: this is work someone has to do, and then keep doing month after month.

  • Where it runs: a server, a cloud service, a computer in the office?
  • Address and domain: how do colleagues reach it, including from outside the office?
  • Certificate and HTTPS: who installs it and who renews it?
  • Uptime: who notices when it stops, and who restarts it?

Data: the database, where it sits, the backups

A business system is worth the data it holds: customers, items, orders, invoices, appointments. That data belongs in a proper database, because a file that gets overwritten will lose it sooner or later. You also need to know where it physically sits: for a European company, “does the data stay in the EU?” carries real weight. You need automatic backups and a way back when a record is deleted by mistake. And there is almost always a concrete first step: bringing in the spreadsheet the company has used for years, with its columns, its duplicates and the same names spelled differently each time. The chat can suggest how to go about it, but actually doing it without losing anything is still up to you.

Users and access: who gets in, who sees what, who changed what

As soon as two people use the program, it needs a login with a password. As soon as five do, it needs roles: the warehouse worker should not see margins, the salesperson should not delete invoices, the owner should see everything. And sooner or later comes the question “who changed this figure, and when?”. Answering it takes an audit log, a trace of every change with its author, time and content. These parts can be written, of course, yet they are also where a slip costs the most: one forgotten check and the data is open to people who should not see it.

  • A password login for each colleague, with passwords stored safely.
  • Roles: who can view, who can edit, who can delete.
  • A record of who did what, which nobody can retouch.
  • What happens when a colleague leaves: their access closes at once.

Maintenance: the change you will need next month

A business system is never quite finished. Next month you will need another field, a new order status, a different report for the accountant. On the chat route, every change starts over: a conversation does not know your program as it is on the server today, so you paste the code in, explain the context again, get the corrected version, copy it back into the right place, and someone has to put it online again. If the changed piece breaks another, you find out when a colleague calls. Then there are updates: the libraries the code rests on change version, and one that works today may stop working tomorrow or have a hole to close. Someone has to follow that too, for all the years the program stays in use.

Security: the risks of the do-it-yourself route

Code written in a hurry and put online by someone who does not do this for a living carries some typical risks. They come from how the code is used, checked and published, whichever assistant wrote it, and it is better to know about them before you go live.

  • Passwords stored in plain text, or logins with no real check behind them.
  • Keys and passwords for other services (payments, courier, email) written into the code, which then gets copied around.
  • Customer data pasted into a conversation “to explain the problem better”: once pasted it has left the company, and someone should decide beforehand what may be pasted and what may not.
  • A server put online and never updated again.
  • No backups, which tends to come to light on the day one is needed.

Building business software with ChatGPT alone: the real path, step by step

To see the difference it helps to line up the steps. Taken one by one they are all doable; the trouble is how many there are and that they come round again with every change. In a small company they usually fall to the colleague who is “good with computers”, in spare moments and on top of the day job. That holds up while the person is around and has time, but once they go on holiday or move on, nobody knows the program any more.

  • You describe in the chat what you need and get the code.
  • You copy the code into the right files and try it on your computer.
  • You choose where to host it and configure the server.
  • You create the database and bring the data in from the spreadsheet.
  • You add logins, passwords and roles, which means more code and more testing.
  • You put it online, with an address and a certificate.
  • You paste the errors into the chat, apply the fixes, put it online again.
  • Next month a change is needed: back to the first step.

What we add

It starts the same way: you describe the program in your own words and the AI builds it. We work with leading AI partners. What changes is everything after that, because here the description turns straight into an application running at its own address, with no blocks of code to sort out by hand. The platform builds, hosts and maintains it; first you settle what the system must do, then how it will be built, and you approve each step. No technical person is needed in the company.

  • Hosted from day one, with the address, certificate and updates handled for you (see Hosting).
  • A database for each project, with backups and a way back to an earlier moment.
  • Data is stored in the European Union, in the Frankfurt region, and is not used to train AI models (see Data security).
  • Colleagues sign in with roles and permissions, and every action stays in an audit log.
  • Keys to outside services stay hidden from the code.
  • To make a change you simply ask for it in the conversation; every version is kept and you can go back when needed.
  • The code is yours: you push it to your own GitHub repository with one button, and the data comes out as a table whenever you want.
  • Your spreadsheet is imported as it is, with its own columns.

When ChatGPT alone is enough, and when it is not

Here is a straight list to help you make up your own mind. If the first few points sound like you, the chat will probably do. If the later ones do, you need a proper program that someone hosts and maintains.

  • The chat is enough: a formula, a macro, a script only you use.
  • The chat is enough: a calculation or conversion you do once and then forget about.
  • The chat is enough: understanding an idea or writing the specification before talking to anyone.
  • You need a program: several people use it, each with their own role.
  • You need a program: it holds customer, invoice or patient data.
  • You need a program: it must work every day, even when whoever built it is on holiday.
  • You need a program: it will change over time, along with the way the company works.

Examples of programs companies describe

These are the systems that most often still live in a spreadsheet or a notebook. Each page describes what the program does, with no made-up customer stories. How to write the description, with an example, is in the guide AI software from a description.

What it costs, and where to start

ChatGPT has its own plans and prices, which we do not repeat here because they change; all the work after the code (server, database, the hours of whoever looks after it) comes on top. With us there is no subscription and no per-user fee: you use credits bought in advance, spent according to the work done, and each active project has a daily hosting fee, while a paused one costs nothing and stays as it was. Packs and prices are on the Pricing page. Signing up is free and needs no card: describe your program and look at the result before you pay. If you would rather not go it alone, we build it together with you: see Services or ask for a quote.

Frequently asked questions

Can ChatGPT build a complete business system?

It can write the code, even much of it. What it does not do on its own is put it online, give it a database, logins and roles, and maintain it over time. Someone has to do that part.

If I already have code made with ChatGPT, can I host it with you?

Yes. A project built elsewhere can be uploaded and hosted here on the same terms as ours, and the code stays yours. See Hosting.

Is AI-CLOUD Italia ChatGPT?

No, and it is not affiliated with OpenAI. It is a platform that builds, hosts and maintains applications from a description, and it works with leading AI partners.

Do I need to know programming?

No. You need to be able to explain what goes in, what has to happen and what comes out. The AI writes the code; running and maintaining it is our job.

Can I import my spreadsheets?

Yes. Tables are imported as they are, with their own columns. Duplicates and different names are sorted out afterwards, inside the program.

What happens when I need to change something?

You ask for it in a new conversation, on the same project. Every change is a version: you see what changed and can go back when needed.

Who owns the code and the data?

You do. The code goes to your own GitHub repository with one button and the data comes out as a table whenever you want. Nothing is deleted automatically.

How much does it cost?

Signing up is free and you do not need a card. After that you pay for what you use, with credits bought beforehand and no subscription, and each active project has a daily hosting fee. Prices are on the Pricing page.

Describe your system and see it running

Sign up at no cost and with no card. Describe the program you need and you will see it online and ready to use before you have paid a thing.