Business growth worldwide is increasingly being influenced by data-driven innovation. In Kenya, for instance, private and public institutions are also gradually turning to generative AI to take more control of their processes and data.

This is particularly evident in the area of application development. With large language models (LLMs) trained on billions of coding parameters, GenAI helps developers by generating quick fixes for repetitive tasks such as boilerplate codes, database operations, and standard UI elements, allowing for a faster and better application development experience.

Will Gen AI replace low-code development platforms?

It is important to note that even before GenAI entered the app development space, low-code and no-code development platforms had been offering a simplified development environment by helping businesses adopt agility in software development life cycles and go-to-market strategies. Consequently, it is not surprising that industry discussions have shifted to whether or not GenAI can completely replace low-code platforms, given its potential and steady uptake. The answer is no.

While GenAI cannot completely replace low-code platforms without human involvement, it may greatly enhance the value that these platforms offer. For instance, based on your prompt, GenAI can easily generate a specific block of code in a specific programming language for a particular functionality. However, what AI can’t do seamlessly is tell you where a snippet should be plugged in, what would change if you tweaked a certain component, or if the generated snippet has the scope for optimisation in view of the desired result.

Google’s Hustle Academy Re-launches with AI Focus to Empower African SMBs

The better option is to select a suitable low-code/no-code platform with extensive GenAI integration and strong LLM capabilities in order to get the most out of everything. A loosely coupled low-code/AI approach may result in technical debt, poorly designed applications, and compliance challenges.

What to keep in mind while using low-code platforms with AI capabilities.

1. Platform maturity: Be careful not to opt for a low-code platform solely for its AI capabilities because the tech itself is still fairly wet behind the ears. Instead, focus on choosing a mature low-code platform with a good breadth and depth of features to support different personas to build scalable custom applications.
2. Privacy and security: Apps built on low-code platforms typically interact with all kinds of data residing within and outside an organisation, depending on the use case. Questions regarding how the data will be used in the context of LLM should be addressed upfront. A platform that has prioritised data privacy and security at its core will focus on extending the same for all of its provisions beyond LLM capabilities.
3. Compliance: Non-compliance is probably the fastest way to bankruptcy. On the one hand, heavy fines are a reality. On the other, it can (and will) drive away customers and prospects, making survival difficult. It’s critical to choose platform vendors that comply with all major regulations from the regions they function in. For example, the platform should actively track the application progress and flag for GDPR non-compliance if the application is designed to be consumed in the EU.
4. Governance: With different developer profiles leveraging the platform to solve real-time business problems, having adequate governance measures at the platform level is important. Adding LLM-based capabilities to the mix further compounds the governance issue because users will inevitably bring in foreign code blocks to build on top of the existing code base.

Approach to adoption

In the near future, we will see all long-term low-code platforms that are reputable train and launch their proprietary LLMs to have better control over the outputs. In addition, we will see contextual domain-centric LLMs deployed with low-code platforms to build industry- or use-case-specific applications at scale.

Just as with the adoption of any new technology, it is not advisable to jump in head-first because of hype or for the sake of it. Rather, the approach should be a conscious effort towards starting small, assessing progress, and scaling.

A low-code platform alone can only do so much to strike a balance between abstraction and control. Offering comprehensive onboarding interventions is one method to bridge this gap.