The synergy between AI and low-code: a look ahead at the future of development

Talking to Armando Jacobus gives us a unique insight into his view on the impact of AI on low-code development. Armando shares his experiences with the rise of AI tools such as Github CoPilot and ChatGPT and how they are changing the software development landscape. At the same time we take an honest look at Mendix, one of the most popular low-code platforms and the one JAM-IT works with.

The power of AI in development

At JAM-IT we build applications in Mendix, one of the most popular low-code platforms. We do not do everything in Mendix, though. Almost always we add elements that we develop the traditional way, by writing code. And that part of our work is changing drastically at the moment, thanks to AI tools such as Github CoPilot and ChatGPT.

To illustrate that, let me take you through how I use CoPilot while writing code.

I ask CoPilot something like: create a registration form and make sure an email is sent to this address after submission. CoPilot first looks at the rest of my code to see how other forms are built, so that everything stays consistent. It then writes the code needed. Of course I still have to check it and perhaps adjust it, but the basis is there.

Do I take over a project from someone else? Then the code sometimes contains functions whose workings I do not immediately understand. I have CoPilot explain those.

CoPilot also makes it far easier for me to refactor outdated applications (which means cleaning up and improving the code). CoPilot can rename variables for more consistency, convert loops into more modern constructions or split large functions into smaller, more manageable pieces.

And once I am happy with my code? Then I ask CoPilot: do you see any mistakes here? And then: would you adjust it for me?

All of this already saves me a few hours of work per week. Across our whole team, the time saved is far greater still. And we are only at the start of these developments.

This video demonstrates how to create your own API to retrieve images, with Copilot as an aid in analysing the API:

Low-code platforms are embracing AI more and more

At the moment these AI tools mainly work well if you develop the traditional way, by writing code. In the low-code platforms you cannot use CoPilot and ChatGPT yet. That is because those platforms are set up very visually, while CoPilot and ChatGPT work on the basis of text.

But the low-code platforms are not lagging behind, of course. At JAM-IT we work with Mendix, and we already see a lot of handy AI-based aids in it. I expect Mendix to make it possible to develop within their platform on the basis of text before long. That is how the full power of AI gets unlocked.

Mendix is going all in on usability and is making everything more and more visual. In the background Mendix works with traditional code too, but that is increasingly hidden behind the visual layer. That is a good thing, and being able to work visually is especially welcome for beginners.

The first time you look at traditional code, your head starts spinning. It is a whole new language, and you understand none of it. The visual side of Mendix makes everything much more intuitive. You see what you are doing. That lowers the barrier. As a more experienced user, though, you can run into challenges. AI tools work on the basis of text, and cannot be used in Mendix. With Mendix it is not yet possible to have a new function generated from a simple instruction, or to rename a large number of elements in one go. Fortunately Mendix also contains plenty of handy aids (some of them AI-based) that do speed up my work.

Unit test in action:

Prospects for the future

I am looking forward to the future. Looking at what has changed in a year, I cannot wait to see where we are a year from now. The future of development is full of promise and innovation. With Mendix as a trusted platform and AI as an integral part of the development process, we can push new boundaries and evolve faster than ever before.