AI coding tools: how I use them, and what every company should check

Lately I have been using AI coding tools in my own work: Claude Code, ChatGPT Codex and Gemini Antigravity. They can write code, fix errors and explain old programs in plain words.

Many company owners now ask me the same questions. “Can AI build our software now? Do we still need developers?” Here is my honest answer, based on my own use.

1. What these tools do well

• First drafts. Screens, forms and tables that used to take days can now be drafted much faster.
• Repeated work. Search filters, reports, import and export of Excel files.
• Explaining old code. Pasting an old Excel macro or an old program and asking “What does this do?” often gives a clear answer.
• Tests and notes. They help write test cases and simple documents, which many teams skip.
• Quick prototypes. You can see a working sample early, before you decide to invest.

2. Where they go wrong

• They can be wrong and still sound sure. The code may look fine and fail on a real case.
• They do not know your business. If you do not explain your process, they will guess.
• Security is easy to miss. A tool may leave a door open that no one noticed.
• Big systems get messy. Without guidance, small changes can slowly break other parts.

3. What I always check by hand

Whatever tool I use, I check these myself:

1. Who can log in, and what can they see. A dealer should never see another dealer’s prices.
2. Where passwords and keys are kept. They must never sit inside the code.
3. What data goes into the tool. I do not paste customer data or company secrets into public AI tools.
4. Real examples. For a quotation system, I test with real quotations and compare the numbers.
5. What happens when something fails. Backups, error messages and a way to go back.

4. What this means for your company

AI tools make good developers faster, so small custom tools are now easier to build. But you still need someone who understands your process and checks the result.

If you are thinking of a customer portal, a quotation system or a catalogue, see our solutions and our case studies. For example, we put more than 60,000 bearings online for PMK Machine Tools, with a quotation request on every product.

In many companies, the Microsoft 365 tools you already pay for can solve the problem without any new software. I look at this first. You can read about it in Power Platform and Power BI.

5. If your own team wants to use these tools

Start small, and set simple rules:

• No passwords, keys or customer data in prompts.
• A second person reviews every important piece of code.
• Keep the code in a proper version system, so you can go back.
• Test before anything goes live.
• Train your people, so that they know both the strengths and the limits.

I run a hands-on program for teams on this topic, called Building with AI Coding Tools. It takes 2 to 4 weeks, and your team builds a small working tool during the program. See the details under corporate training.

In short

AI coding tools are a good help, but they work best in the hands of people who understand the business and check the result. If you are not sure whether they suit your company, talk to me. I will tell you honestly.

Have a question about your own situation? Message me on WhatsApp and I will reply personally.