Low-code software development assisted by AI

Artificial Intelligence (AI) has been in the spotlight lately. This is partly due to a number of technological innovations recently unveiled by the American tech company OpenAI. Last year, OpenAI drew attention to itself with the introduction of Dall-E, a generative AI model that can create unique images based on instructions in human language. A real hype followed when OpenAI launched a new version of the intelligent chatbot ChatGPT on 30 November and made a free version available online. Within a few days of the launch, more than 1 million users had already signed up for ChatGPT.

OpenAI's new AI applications have sparked many people's imagination about the possibilities of artificial intelligence. Following the enormous popularity of ChatGPT and Dall-E, Microsoft announced in January that it is investing billions in OpenAI. This has created high expectations about the use of AI in existing software such as the Bing search engine and MS Office.

What does this leap in AI technology mean for software development?

It has long been no secret that developers are scarce and that this situation will continue for some time, while the need for new software applications keeps growing. Can AI offer a solution? Can generative AI models such as ChatGPT take over the work of software developers?

The holy grail may well be that anyone can be a programmer and that you can ask an AI in human language to generate an application. We are still a long way from that. Even so, the use of AI in software development is not a distant prospect, because it is already happening in modern low-code/no-code tools such as Mendix and Power Apps. In this blog, you'll read which AI applications are already available in these tools and which innovations we can expect.

AI-driven tools in low-code platforms

We saw the first applications of AI and ML (machine learning) in low-code tools as early as 2018, when Mendix introduced the smart assistant Mendix Assist. Mendix Assist helps developers model application logic in so-called microflows by continuously suggesting improvements and adjustments as the next step. To do so, Mendix Assist uses historical, anonymised data based on the many application models customers have developed on the Mendix cloud platform. The Mendix Assist AI model recognises patterns while you are modelling and gives you contextual recommendations for a next step, and if you accept a suggested recommendation, it can carry it out for you. That way, you not only build better applications, you also develop them faster than if you had to do it all yourself.

In the years since its introduction, Mendix Assist has been extended with new features, including for creating pages (screens) and automatically generating validation flows. Besides speeding up the development process, Mendix Assist now also helps optimise application performance. The Mendix Assist Performance Bot checks an application on all kinds of aspects during development and detects potential design flaws in the data model, the microflows and the pages of an application. Detected problems in the application design are clearly explained to the developer and can then be fixed automatically by Mendix Assist.

In May last year, Microsoft launched Express Design for Power Apps. With Express Design, an existing wireframe (design) from Figma, an image of a paper form and even a screenshot of a whiteboard can be converted into an app automatically. The Power Apps cognitive AI model recognises the structures in an image and can translate them into a working user interface for a new application. The model is even able to create a data model in Dataverse based on the image.

At the last Ignite event in October last year, Microsoft presented an AI innovation for Power Automate that gives us an idea of what to expect in the near future. It showcased a preview feature, Create a flow from a description, which lets you describe what you want to automate in a command bar. Power Automate then translates the text you enter into code. Based on that code, a flow is automatically modelled for you with the activities and parameters needed to carry out the requested automation. Power Automate uses OpenAI Codex for this, which, like ChatGPT, is a generative AI model, in this case for generating code.

Embedding AI functionality in low-code apps

Another trend we see in low-code software development is the integration of AI functionality into applications. Think of connecting an application to standard cloud AI services for automatic image recognition or converting audio to text. These services can be linked to a low-code tool fairly easily. In addition, many companies are developing custom AI models for specific applications, such as process optimisation, fraud detection and predictive maintenance. Until recently, embedding these kinds of AI models in low-code applications was difficult, partly because complex APIs and REST services were needed to call the models from a low-code tool. Recent innovations in Mendix and Power Apps can solve these problems.

With AI Builder, Microsoft has for some time offered the ability to create AI models and use them within Power Apps. In AI Builder, you can choose to use a prebuilt model or build a model yourself in a low-code way (without programming). AI Builder offers templates for various model types, including category classification, document processing and predictions. Microsoft recently extended AI Builder's capabilities, so you can now import custom ML models developed with the Azure Machine Learning platform into AI Builder. Once an AI model has been created in AI Builder, it can be called in Power Apps using Power Fx. This is a low-code formula language from Microsoft that works in much the same way as Excel.

Mendix is more or less following Microsoft's lead with the new Machine Learning Kit, announced at the end of last year and due to be officially released very soon. This too should make it possible to add a custom ML model to an app, making it much easier to call the model within Mendix.

What can we expect?

So far, it is mainly smart AI-driven features in low-code tools that make developing applications easier and result in better software being produced with these tools. AI acts as a kind of co-pilot, taking tasks off the developer's hands and preventing the developer from making design mistakes. In the near future, I expect more of these kinds of applications, where developers are helped not only with building applications but also with other aspects of software development, such as automated application testing.

With the recent technological developments in generative AI, I expect we will see new AI applications in low-code tools this year. With the investment in OpenAI that Microsoft has announced, it's quite possible that a smart AI chatbot will be integrated into Power Apps and Power Automate. The first signs of this can already be seen in preview features such as Power Apps Ideas and Create a flow from a description in Power Automate. These make it possible to ask, in human language, for a specific part of an application to be modelled. Now that's what I call next-level low-code app development!

Will AI make software developers redundant?

Will flesh-and-blood software developers soon be made redundant by AI? I don't believe so. If code is generated automatically by AI, you as a developer need to be able to review the code and adjust it where necessary. We cannot rely on the output of generative AI models always being correct, or on it matching the client's intentions exactly. Moreover, software development consists of much more than code. Talking to users and translating their wishes into functionality that can actually be built remains human work. The expert role of an experienced developer or tool specialist therefore remains essential in software development.

Want to know more about low-code?

Would you like to know more about the possibilities of low-code tools, or do you have a specific challenge where a low-code app could provide the solution? We are happy to advise you and think along about possible applications of low-code for your specific situation. Get in touch with one of our account managers for an introductory meeting.

Schedule an appointment

Want to get started as a low-code developer yourself?

The demand for low-code software is only increasing. That is why E-mergo is looking to strengthen the team that builds solutions for customers with Mendix and Power Apps. Do you see yourself as a future low-code developer? Don't hesitate and check out our vacancies on the website via the button below.