AI-CLOUD.GR

We work with leading AI providers.

ClaudeGeminiOpenAIAWSVertex AIDatabricksDeepSeekCerebrasQwenStability AI

Guide

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

Small business owners ask us this a lot: “ChatGPT writes code, so why pay for software?” It is true that it can write code that works. Getting from that code to a program your company relies on every day is a longer road, though. This guide covers what ChatGPT does well, what still has to happen after the code is written, and when it is enough on its own.

The short answer: yes, but code is not yet a program

A chat assistant gives you code or text inside a conversation. For a small script you run once, a spreadsheet formula or a tool only you will use, that is often enough. A business program asks for more. It has to run somewhere around the clock, keep your data safe, know which colleague is signing in and what they are allowed to see, and change when your business changes. Some versions can run or preview small pieces of code inside the conversation, which is handy for a quick test, though it is still a long way from a hosted system the whole team works in. Most of the effort sits in the gap between having the code and having a working program.

What ChatGPT does well

To be fair, ChatGPT, the chat assistant from OpenAI, is a very good tool for thinking a problem through, getting something explained and making a start. For many of a company’s smaller needs it is exactly what you want, and there is no reason to look further.

  • It explains in plain words a technical idea, an error message or code someone sent you.
  • It writes a spreadsheet formula or a small script that renames files or cleans up a table.
  • It sketches what a data table could look like, with its columns and how a customer links to an order.
  • It drafts code for a simple page or form that a developer can then pick up.
  • It helps you write the specification (what goes in, what has to happen and what comes out), which is the most useful step before any program.

Hosting: where it runs and who keeps it on

Code from a conversation doesn’t run anywhere until someone puts it on a server. That means choosing a provider, uploading the files, setting up the address and the domain, and adding an HTTPS certificate that keeps renewing. You also need a plan for the Saturday night when the program goes down. A chat won’t alert anyone when the app stops, so you usually find out from a customer’s phone call. On top of that, someone has to keep the operating system and packages underneath it updated without breaking anything.

  • Where it runs: a server or hosting service, with an account and a bill someone has to keep an eye on.
  • Address and domain: DNS settings that are usually done once and forgotten until the day they break.
  • HTTPS: the certificate has to renew, otherwise visitors get a browser warning.
  • Uptime: someone has to notice that something has stopped and start it again.

Data: a database, backups and your spreadsheet

A business program depends on its data, meaning customers, items, orders and invoices. Code from a conversation often keeps them in a file or in memory, which is fine for a test and nowhere near enough for daily work. You need a real database, someone to set it up, and a clear answer about where the data physically sits, because GDPR asks that and so does every larger customer. Then there are backups. Having them means little if nobody has ever tried restoring one. Finally, the spreadsheet you have kept for years has to move into the new program with your own columns and names.

  • A database that copes with many users at once, rather than a file on a disk.
  • A known home for the data, meaning which country, which provider and under which contract.
  • Backups that run automatically, and a restore that has actually been tested.
  • Importing your existing spreadsheet without anyone having to retype it.

Users and access: who signs in, what they see, who changed what

The moment two people use the program, it needs a password login. By the time five do, it needs roles as well: sales shouldn’t see margins, an outside partner shouldn’t see salaries, and the accountant can look without changing anything. On the day a number turns out wrong, you will want to know who changed it and when. A team expects all of this within the first week, and it is among the hardest parts to get right, because a mistake in login or permissions usually stays hidden until someone else stumbles on it.

  • A login with properly stored passwords, password reset, and access removed for anyone who leaves the company.
  • Roles and permissions checked on every request to the server, since a hidden button protects nothing.
  • An audit log recording who did what and when, which you will be glad of in any internal review.

Maintenance: the change you will need next month

No business program stays the way it was on the day it was delivered. Next month you will want a new field, a report or a reminder to customers. A chat conversation, however, doesn’t remember your system as it stands today. For each change you paste the code back in, explain again how it is put together, get new code and work out exactly where it belongs. Then someone has to deploy it without disturbing the data already there. Some changes you never asked for, too: a library the code depends on releases a new version, and something that worked yesterday stops working today.

  • Each change is another round of copy and paste, with a risk of breaking something that worked.
  • Deploying the new version to the server is a separate job from writing it.
  • Without kept versions, getting back to “how it was yesterday” is not always possible.
  • Library and security updates keep coming for as long as the program is in use.

Security: passwords, secrets in code and the data in the chat

The do-it-yourself route also carries risks you won’t see on day one. They have nothing to do with any particular tool. They come from the program being put together without anyone who knows what to look for reviewing it. Code can run perfectly well and still have security holes, and people usually discover that after the damage is done.

  • User passwords stored readable instead of as a one-way hash.
  • Keys and passwords for other systems (payments, courier, accounting) written into the code, ending up in a file, an email or a paste.
  • Customer data pasted into a conversation so the tool can “see” the problem, which puts it outside any company system with roles and an audit log. Before you do it, check the terms of the provider you use and your own GDPR obligations.
  • Forms with no protection against automated submissions, and connections without TLS everywhere.

The realistic do-it-yourself path, step by step

To give you a clear picture, here is what building a business program yourself with a chat assistant looks like in practice. None of the steps is impossible. The real question is who in your small company will do all of them, every time, on top of their own job, and what happens when that person leaves or goes on holiday.

  • You describe what you want and get code.
  • You copy it into files on your computer and try to run it.
  • You pick a hosting provider, open an account, set up a domain and HTTPS.
  • You set up a database and connect the code to it.
  • You deploy the app and fix the errors that only show up on the server.
  • You paste those errors back into the conversation, get new code and deploy again.
  • You repeat all of it for every change, every new user and every update.

What the AI-CLOUD Greece platform adds

Our platform is not ChatGPT and not a version of it. On it you describe the program in Greek, the way you would in a conversation, and the AI builds it. The difference is that you end up with an application already running at its own address, so there is no code left for you to deploy. Everything described above as work “after the code” happens in the same place as the build itself. More on the Platform and Hosting pages.

  • Hosting from day one, with the address, certificate and updates handled without any effort from you, plus your own domain if you want one.
  • A database inside the project with automatic backups, and your spreadsheet imported as it is.
  • Data is kept in the European Union, in the Frankfurt region (see Data security).
  • Colleagues sign in with roles and permissions, and every change stays in an audit log.
  • Keys for other systems are stored separately and never appear in the code.
  • A change takes a new conversation with no new order, and each change is kept as a version you can roll back to.
  • The code is yours and goes to your own GitHub repository.

When ChatGPT alone is enough, and when it is not

Not every need calls for a platform. If yours matches the first cases on the list, carry on with the chat and you lose nothing. If it matches the later ones, the code will be the smallest part of what you need.

  • Enough when only you use it, once or rarely.
  • Enough when it is a formula, a macro, a small script or a piece of text.
  • Enough when the data is neither personal nor critical to the business.
  • Not enough when two or more colleagues work in it, with different permissions.
  • Not enough when it has to be available every day, including to customers or partners outside.
  • Not enough when it holds customers, orders, invoices or health data.
  • Not enough when you already know you will want changes every so often.

Examples of programs companies describe

Most companies start from something they currently keep in a spreadsheet, a notebook or a colleague’s head. Below are some of the systems people describe to us most often, and each has its own page on what it does and what it doesn’t. For how to describe yours, with an example, see the guide AI software from a description.

What it costs and how to start

ChatGPT is paid for on whatever terms its provider sets. On the do-it-yourself route, though, that is rarely the biggest expense, because hosting, the database and above all the hours of whoever sets it up and maintains it come on top. With us there is no subscription and no per-user fee. You top up prepaid credits that are used according to activity, each active project has a daily hosting fee, and a paused project costs nothing. Packs and their prices are on the Pricing page. Your account opens free of charge without card details, so you can describe the first task and see the result before paying anything. If you would rather we build it with you, see Services or send a quote request.

Frequently asked questions

Can ChatGPT build a whole program for my company?

It can write much of the code. The program also needs hosting, a database, user logins with roles, backups and someone to maintain it, though. Those don’t come out of a conversation; a person or a platform has to set them up and look after them.

Can I build software with ChatGPT without knowing how to program?

A small tool for yourself, perhaps. A system for the whole team is harder, because errors, deployment and security need someone who understands what the code does. On our platform you never see the code. You describe what you want, check the result and ask for corrections in plain words (see vibe coding in Greek).

Is AI-CLOUD Greece ChatGPT?

No. AI-CLOUD Greece is not ChatGPT, nor a version of it. It is a platform from 3D Code EOOD for building and hosting applications from a description. The language model providers we use are listed on the Data security page.

I already have code from ChatGPT. Can I host it with you?

AI-CLOUD Greece also accepts for hosting a project built elsewhere, by you or by another team, on the same terms. You will find more on the Hosting page, or write to us and we can go through it.

Is it safe to paste customer data into an AI chat?

That depends on the provider’s terms and on your own GDPR obligations, so check them before you do it. Either way, anything you paste into a conversation sits outside your company’s systems, with no roles and no audit log.

What happens when I need a change months later?

With a chat, you explain the system again, paste the code in, and someone deploys the new version. On the platform you describe the change just as you described the original program, and it is saved as a new version that goes straight to the application already running.

Who owns a program built on the platform?

You do. The code goes to your own GitHub repository and the data can be exported at any time. Nothing is deleted automatically, and if your credits run out the project simply pauses without losing anything.

Describe the program and we keep it running

Open an account at no charge and with no card, write down in Greek the first task you want solved, and see what gets built before you spend a single euro.