AI and low/no code: What they can and can’t do together

Artificial Intelligence (AI) is in the fast lane and driving  toward mainstream enterprise acceptance, but, at the same time, another technology is making its presence known: low-code and no-code programming. While these two initiatives inhabit different spheres within the data stack, they nevertheless offer some intriguing possibilities to work in tandem to vastly simplify and streamline data processes and product development.

Low-code and no-code are intended to make it simpler to create new applications and services, so much so that even nonprogrammers – i.e., knowledge workers who actually use these apps – can create the tools they need to complete their own tasks. They work primarily by creating modular, interoperable functions that can be mixed and matched to suit a wide variety of needs. If this technology can be combined with AI to help guide development efforts, there’s no telling how productive the enterprise workforce can become in a few short years.

Intelligent programming

Venture capital is already starting to flow in this direction. A startup called Sway AI recently launched a drag-and-drop platform that uses open-source AI models to enable low-code and no-code development for novice, intermediate and expert users. The company claims this will allow organizations to put new tools, including intelligent ones, into production quicker, while at the same time fostering greater collaboration among users to expand and integrate these emerging data capabilities in ways that are both efficient and highly productive. The company has already tailored its generic platform for specialized use cases in healthcare, supply chain management and other sectors.

AI’s contribution to this process is basically the same as in other areas, says Gartner’s Jason Wong – that is, to take on rote, repetitive tasks, which in development processes includes things like performance testing, QA and data analysis. Wong noted that while AI’s use in no-code and low-code development is still in its early stage, big hitters like Microsoft are keenly interested in applying it to areas like platform analysis, data anonymization and UI development, which should greatly alleviate the current skills shortage that is preventing many initiatives from achieving production-ready status.

Before we start dreaming about an optimized, AI-empowered development chain, however, we’ll need to address a few practical concerns, according to developer Anouk Dutrée. For one thing, abstracting code into composable modules creates a lot of overhead, and this introduces latency to the process. AI is gravitating increasingly toward mobile and web applications, where even delays of 100 ms can drive users away. For back-office apps that tend to quietly churn away for hours this shouldn’t be much of an issue, but then, this isn’t likely to be a ripe area for low- or no-code development either.

AI constrained

Additionally, most low-code platforms are not very flexible, given that they work with largely pre-defined modules. AI use cases, however, are usually highly specific and dependent on the data that is available and how it is stored, conditioned and processed. So, in all likelihood, you’ll need customized code to make an AI model function properly with other elements in the low/no-code template, and this could end up costing more than the platform itself. This same dichotomy impacts functions like training and maintenance as well, where AI’s flexibility runs into low/no-code’s relative rigidity.

Adding a dose of machine learning to low-code and no-code platforms could help loosen them up, however, and add a much-needed dose of ethical behavior as well. Persistent Systems’ Dattaraj Rao recently highlighted how ML can allow users to run pre-canned patterns for processes like feature engineering, data cleansing, model development and statistical comparison, all of which should help create models that are transparent, explainable and predictable.

It’s probably an overstatement to say that AI and no/low-code are like chocolate and peanut butter, but there are solid reasons to expect that they can enhance each other’s strengths and diminish their weaknesses in a number of key applications. As the enterprise becomes increasingly dependent on the development of new products and services, both technologies can remove the many roadblocks that currently stifle this process – and this will likely remain the case regardless of whether they are working together or independently.