The idea of useful applications created without writing a line of code certainly isn’t a new idea to take flight, but it is reaching cruising altitude. The makers of low-code platforms hope AI-powered natural language chatbots will get low-code development to its full potential. Still, some say that adding additional natural language AI capabilities to chatbots isn’t the right way for enterprises to consider their low-code efforts.
The history of creating useful applications without human programmers goes back to the 1970s with the concept of fourth-generation programming languages and later rapid application development tools. And in 1982, James Martin’s book, Application Development Without Programmers, was published. Martin foresaw a computing world without programmers. But it’s only been since the advent of modern low-code development platforms, and citizen developers are bringing that vision closer to reality.
The term low-code didn’t exist until 2011 when Forrester Research first used it in one of its reports about new productivity platforms for custom application development. Since then, adoption has been slower than some hoped, but it’s also been steady and accelerating. Recently, low-code platform providers hope incorporating AI into their platforms will make developing applications more accessible to non-developers and help professional developers become more productive.
Today, low-code tool adoption among professional developers is nearly ubiquitous. “Most professional developers use at least some low-code. The statistics are staggering, upwards of 90%,” John Bratincevic, principal at Forrester Research, says. “It seems everyone who is a developer is using low-code, at least to a small extent,” he adds.
The market for low-code services and platforms is soaring. According to Research and Markets, the worldwide low-code market reached $10.9 billion in 2019 and is expected to reach $187 billion by 2030. Still, despite the low-code label, developing applications at scale with low-code platforms — especially in enterprise environments — remains a challenge.
Low-Code Meets AI
To bolster their toolsets, low-code platform vendors are currently introducing their products and services with generative pre-trained transformer models (GPT models) to make their tools more productive to professional developers and more accessible to citizen developers. Dev Nag, founder and CEO at QueryPal, says companies are starting to rely on AI enhancements. “These platforms are where AI coding makes the most sense, as they represent the most common, tried-and-true applications with a ton of training data,” he explains.
For instance, this week, integration platform-as-a-service provider Jitterbit announced the addition of AI assistants to its Harmony platform. The platform provides customers data integration, workflow automation and low-code application development capabilities. Jitterbit says it’s now integrating AI into bots across its Harmony platform: App Builder AI Assistant, Connector AI Assistant, and the AskJB chatbot (which will field Jitterbit’s documentation questions). The AI assistants are currently in beta.
Jack Yao, CIO at workers compensation insurance provider MEMIC, said in Jitterbit’s statement that, as part of the beta program, the AI capabilities in App Builder AI assistant have helped develop their claim reserving and producer commission applications. “Our experience interacting with the App Builder AI assistant was a critical efficiency enabler. The AI prompts and order of operations were logical and intuitive saving us weeks in creation versus our manual process,” Yao said.
In June, with its release of Oracle APEX 24.1, Oracle added AI capabilities to its low-code development platform, including a conversational companion APEX AI Assistant, AI-Assisted App Development, AI-Assisted SQL Authoring, AI-Assisted Debugging, and more AI enhancements, including integration with several third-party AI providers and the ability to create application blueprints from natural language prompts.
Other low-code platform providers, such as Mendix and OutSystems, have recently added AI capabilities, too, such as Mendix’s AI-assisted development bot, MxAssist, and OutSystems with external AI integrations and 25 million anonymized patterns that can be built into applications and its AI mentor that guides low-code developers.
Low-Code for Everyone, Everything
Many enterprises have not found how low-code strategically and tactically fits within their organizations. For instance, Nag says low code works ideally for greenfield ideas. “Like any prototypes, the AI-generated code is often easier to work for a brand-new idea than for all of the revisions and enhancements that follow. AI tends to work best when it doesn’t require any implicit context, as in a brand new prototype. Once you have a working application with user behavior and unspoken expectations, AI has less firm footing and requires more human guidance,” Nag says.
Still, Nag and other experts contend that low-code tools provide citizen developers with little experience in developing practical applications. “These tools can certainly help non-developers develop workable, usable applications – as long as the applications are relatively simple and look similar to other apps,” Nag says. “If you’re building the world’s hundredth alarm clock app or workout routine app, AI can generate very reasonable code the first time out. But if the app is more complicated, or somewhat novel, non-developers won’t be able to guide the AI usefully, and it’s likely to cause far more frustration than working directly with an experienced developer from the start,” he claims.
Still, businesses have embraced low-code tools to varying degrees. “There are a few dozen companies that are using low code at a massive scale,” Bratincevic says. “There are probably thousands of organizations that are now at a decent low code development scale, where they have 1,000 or 2,000 business people making stuff, and they’ll have a few thousand applications. And then there are smaller companies with dozens of apps developed with low-code,” he adds.
Bratincevic questions the value of using natural language user interfaces to create low-code apps. “That’s not the right way to think about engineering. The right way to think about low-code is that you know that you need to create a piece of software, and at the end [of development], you have useful software running. How do we shorten the distance between the two points as much as is humanly possible while maintaining quality,” he asks. The correct strategy, he explains, is to assemble a platform of components to accomplish generating bigger and bigger pieces of software that get from idea to working software as quickly as possible,” he says.
“Thinking like that is very different than working with every software vendor adding little chatbots to figure out how to charge you for the modeling they’re hitting behind the scenes,” Bratincevic says.
Still, as organizations learn how to get more out of their low-code platforms from professional developers and how to manage citizen development across their businesses effectively, the value of these low-code platforms will grow. “The impact of AI coding on enterprise app development will mean greatly accelerated prototyping, a democratization, and thus an explosion, of simple custom app development, and a shift in the daily lives of internal developers, who will be able to spend more time on novel problem-solving and complex logic,” Nag says.
