Practical AI
The Road to AI Generated Apps is Paved with Low-Code/No-Code
Despite much recent conversation about low-code/no-code (LCNC) being replaced by AI-generated applications, citizen developers should be in no rush to abandon their favorite LCNC tools. It would seem that the developers of those platforms are carefully navigating the path from point-drag-click to natural language requests.
Even the Leader is Stepping Carefully
Last month, OutSystems Chief Product and Technology Officer Luis Blando announced the launch of a new feature to their Developer Cloud that he says, “represents a vison of the future of software development.”
“OutSystems Mentor is the industry’s first GenAI-powered digital worker that supports the entire software development lifecycle (SDLC): from design to delivery to ongoing maintenance,” Blando added.
That’s no mean claim. Even Blando acknowledged that this was a “mouthful.” He goes on to explain that users can simply describe their need in natural language or submit a full requirements document that describes the application in detail. Ultimately, he said, “Mentor stands out because it doesn’t just rely on user prompts—it leverages all the information you provide in requirement documents, combined with the context of your environment, to deliver accurate and relevant recommendations and outputs.”
AI-generated software applications made simple.
Not So Fast
In OutSystems’ “The State of Application Development Report,” in a section regarding “GenAI assistance: A matter of trust?” a key consideration is very clearly stated: “An error introduced by GenAI-produced code may have a stronger negative impact in an external app vs. an internal app because of its effect on customer experience. Widely used GenAI tools like ChatGPT bring the most significant risks. According to a recent study, these tools generate accurate code no more than 65% of the time—with accuracy rates for some tools falling as low as 31%.”
They also pointed out that, “Additionally, there is widespread concern that without a governing platform or a defined, model-driven abstraction layer, code written with GenAI can quickly accumulate technical debt, making code more difficult to scale and maintain.”
Insight From the Top
OutSystems’ Founder and Chief Executive Officer Paulo Rosado provided more insight into the progress of “the GenAI disruption” into the world of low-code development, recalling, “We witnessed the quick adoption of GenAI-powered “copilots” that enhance the productivity of teams by helping them generate code faster. Two years later, GenAI has become part of the software development process across the industry.”
Explaining how there are “no free lunches” in software development, Rosado explained,” One issue is that the quality of code has become increasingly harder to control. Copilots enhance team productivity by helping developers generate code faster, but they're also a driver of technical debt,” which describes an implied future cost of additional work required as a result of initially choosing a more expedient development solution versus a more robust one. AI tends to generate more and much longer lines of code that developers will later find difficult to decipher, requiring more tools to make sense of the code. Rosado also pointed out that “you’ll need different tools to ensure critical aspects like explainability, governance, and the security rules and policies that keep generated code under control.”
Not Fully Ready for Prime-Time
Based on a cursory scan of the various LCNC platform providers, it would seem that OutSystems is correct in their claim that Mentor is the first AI “digital worker” designed to help citizen developers produce efficient, useful applications.
That notwithstanding, OutSystems also provides a service to its marketplace and customers in forewarning them to keep in mind the fact that AI-augmented LCNC is brand-new, and still requires more diligence, more vigilance, and a stepwise approach to progress.
Though it is the first of its kind, OutSystems’ Mentor is clearly not going to be the last bespoke AI model developed to support citizen developers in creating the applications their organizations need with less involvement of expensive software development resources. Hopefully, these future entries will learn from the pioneering leadership shown here.
About the Author
Technologist, creator of compelling content, and senior "resultant" Howard M. Cohen has been in the information technology industry for more than four decades. He has held senior executive positions in many of the top channel partner organizations and he currently writes for and about IT and the IT channel.