Is AI-enabled development becoming the new norm? Here's how it looks. At Sweet Mustard, we are closely following the evolutions and exploring the possibilities. We'd like to give an insight into how we at Sweet Mustard have implemented the basic steps. Jos De Berdt, Front-end architect & AI inspirator, talks about how we're approaching it and what we've learned so far.

Follow along to discover what AI can do for other disciplines within software development, such as design, analytics and testing, and what AI agents can accelerate in the future.

What makes AI-enabled development interesting?

Personal code assistant

AI tools act primarily as personal code assistants. It is an example of how AI will not replace your job, but rather make your work easier and more efficient. You can outsource repetitive tasks so you can focus on more complex challenges.

Lower technical debt

AI tools can also add value within the broader picture. Also when it comes to technical debt. This often remains a problem due to tight deadlines and a lack of documentation. And let that be precisely where AI excels. Whereas updating and reacting to code is not the favorite task of most developers, AI is very good at it. The result is less technical debt and a better-maintained codebase.

Fewer errors

AI tools are strong at looking for and fixing errors, which reduces the chance of bugs.

Our approach to implementing AI tools

1. Start an AI workgroup

We started an AI workgroup in which any interested colleague could volunteer. Everyone chose an AI tool to experiment with, and we discussed the findings in two-weekly knowledge sessions. Participants got to know the tools inside out and then passed on their knowledge to other developers who also wanted to jump on the AI wagon.

2. Write clear guidelines

Important when working with AI tools - or new technology in general - are good guidelines. Some examples?

AI is a tool, not a replacement. As a developer, you always remain responsible to control everything.

Always protect the privacy of customers and users

Be careful with sensitive company information.

Want these guidelines too? Send us a message and we'll be happy to provide them to you.

3. Roll-out to the rest of the team

Meanwhile, AI tools are widespread within Sweet Mustard. Our developers are given a limited list of AI assistants to choose from. We host a session where we explain how they work and discuss guidelines, and share the necessary links so they can stay up-to-date and further explore what's possible. That way, our developers make truly informed choices.

Our set-up and tools

AI in development?

With a tool like LMstudio, you can easily set up a private AI. All you need is a Hugging Face account and then Code Llama model, for example, can be used. Private AI ensures that all data stays local to your computer, so no worries around privacy.

You can enhance the development experience by connecting the AI to VSCode via the Continue extension for seamless integration and efficient work.

Our favorite tools:

Our favorite tools is GitHub Copilot, a powerful AI assistant that speeds up our coding experience. Perhaps the most well-known name in the industry. It helps us with smart code recommendations and full feature generation, allowing us to work faster and more efficiently.

At Sweet Mustard, we are already excited: we don't oblige anyone, yet 80% of our developers use AI tools to develop. So we are convinced that other companies can reap the benefits as well!

What to watch for when choosing and launching AI tools

There are a lot of things to consider when choosing your (set of) AI tools. Here are some things that can help you make your choice:

• Language support: choose a tool that is compatible with the programming languages you work in.

• Integration: make sure the tool integrates seamlessly with your existing tools and workflows.

• Efficiency: how user-friendly is the tool? Is it easy to learn and quick to deploy?

• Privacy: what happens to the data you put into the tool? Does everything stay local or is it sent to remote servers?

• Context awareness: a good tool takes into account the project context and generates answers that fit your code style and structure.

• Affordability: free models can be a good option. Depending on your needs, choose local, on-premise or self-hosted solutions.

• Maintainability: maintain 20 tools with licenses? Not ideal. Make targeted choices and focus on what really works.

• Change management: change is difficult, even for developers. Provide guidance, because people stick to their familiar way of working.

Want to know more or would you like our guidelines?

Did you know that our colleague Jos gives presentations on how to integrate AI tools within your development team?

Send a message if you are interested in this presentation, or if you would like to receive our guidelines. After all, at Sweet Mustard we believe that we all get ahead by sharing knowledge!