Blog · AI

AI in shop development: where Claude Code helps – and where it doesn't

AI is part of our everyday development work. An honest look at what it does well, where it gets things wrong and how we deal with that.

For some time now, we at justSelling have been working with Claude Code, an AI tool for software development from Anthropic. It reads code, writes drafts and tests and takes over routine work. Here is a field report – with what works well and what you need to keep an eye on.

Where AI really makes us faster

Understanding existing shops. Old code, unknown extensions, missing documentation: an assessment used to take days. Today, AI gives us an overview in a short time – which modules exist, what they do, where the dependencies are. This helps especially with a Magento 1 migration.

Routine work. Recurring building blocks, integrations with ERP or shipping, frontend adjustments: AI creates a draft, the developer gets it into shape. The gain isn't that AI "programs" but that the developer spends less time typing.

Tests. Automated tests are created along the way – including where they used to be skipped for lack of time. More tests mean fewer errors in operation.

A second pair of eyes. In code reviews, AI finds things you overlook on the third read.

Documentation and translation. Technical documentation grows with the code, and texts for multilingual shops are produced faster.

Where AI makes mistakes

AI makes mistakes – and often sounds very convincing while doing so. Typical examples:

  • It invents functions or settings that don't exist.
  • It solves a problem, but not the one that was actually meant.
  • It misses side effects in parts of the system it doesn't know.

That's why our rule is: what AI delivers is a suggestion, not a result.

Our rules

  1. People decide. Every change is reviewed, tested and owned by an experienced developer – just as before.
  2. Data protection and clear agreements. We use AI only within applicable data protection law. How and with which data we use AI in a project is agreed with the customer beforehand and documented.
  3. Open about risks. We say clearly where AI is a risk and where it isn't the right choice.

What this means for our customers

Faster assessments and quotes, shorter implementation times for routine work, more tests – and still the same contact person and the same responsibility as always. The time saved goes into what makes a shop special.

Want to know how AI can help in your project – and where it can't? Get in touch.

Questions about this topic?

We're happy to talk about your project. Michael Schäfer, phone +49 2802 7079510