AI writes enterprise code fast. Cleaning it up is another story entirely

AI-generated code is accelerating how quickly applications can be built, but creating reliable, enterprise-grade systems remains far more challenging.

As organizations embrace rapid prototyping, many are finding that speed without governance can quickly lead to technical debt and maintenance challenges. Research shows that vibe coding and spec-driven development rarely deliver production-ready enterprise software, with promising prototypes stalling due to issues with functionality and security, according to Medhat Galal (pictured), senior vice president of engineering at Appian Corp.

“It looks great and then they run into one problem after the other. If they can make it functional, they can’t make it secure. When they make it secure and the model changes underneath you, then it changes all over again and then you’re back in maintenance mode,” Galal said. “You end up being a big maintainer of AI systems instead of building the application.”

Galal spoke with theCUBE’s Dave Vellante and Alison Kosik at Appian World 2026, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed the gap between AI-generated code and enterprise-grade systems, as well as the growing risks of technical debt and reliability challenges. (* Disclosure below.)

AI-generated code and the realities of enterprise integration

There’s a reason enterprise software projects take years, not weeks. Generating Java or Python code with tools such as Grok is only the first step — that code still has to be integrated with data, authentication and authorization systems to function securely in a real-world environment, according to Galal. Each layer of functionality introduces its own level of complexity.

“To make two systems talk together — integration is another abstraction layer above them. Then you have to go into orchestration and business rules,” Galal said. “How do you codify these? Then you have to talk about the user interface. How are you going to build those? All of these are different abstraction layers.”

Despite early excitement with AI-driven development, building a stable system is proving to be far more difficult in practice. What starts as rapid progress often turns into a long tail of issues, with more time spent fixing AI-generated mistakes than developing the application itself, according to Galal. It is a problem Appian says it was built to solve — anchoring AI inside governed processes to reduce rework and technical debt, with guardrails that constrain what agents can do and when. The bottom line is blunt: AI development is not the shortcut it appears to be.

“One of my favorite T-shirts Appian has ever had is, ‘I write code so you don’t have to,’ because it’s a really hard job,” he said. “AI is just as hard.”

Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of Appian World 2026:

(* Disclosure: TheCUBE is a paid media partner for Appian World. Neither Appian, the sponsor of theCUBE’s event coverage, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Photo: SiliconANGLE