AI in application development needs more than speed to succeed

The acceleration of AI in application development has reached a new peak — but many organizations are learning the hard way that tooling alone won’t solve deep-rooted inefficiencies. While AI agents and generative platforms promise to help developers move faster, the conversation is increasingly shifting from experimentation to operational maturity.

The reality is clear: AI can enhance productivity, but it can’t replace the foundational capabilities required to build, secure, and operate software at scale. As developers face pressure to deliver new applications faster — often on an hourly basis  — the success of AI integration hinges on how well teams implement automation, maintain infrastructure discipline and support developer platforms with clear accountability.

In the latest episode of theCUBE Research’s AppDevANGLE podcast, Richard Seroter (pictured, right), chief evangelist at Google Cloud, talks with SiliconANGLE and theCUBE’s Paul Nashawaty about why AI alone won’t modernize your applications.

“The AI works for me,” he said. “I don’t work for the AI. That’s the way we’ve got to have our mindset, and there’s still ownership that we have to take.”

AI in application development requires operational maturity

Adopting AI to accelerate application development is no longer optional — it’s becoming a default expectation. But without solid pipelines and DevOps discipline, AI can amplify existing inefficiencies instead of fixing them, according to Seroter.

“You have to figure out infrastructure automation,” Seroter said. “You probably need a mature [application programming interface] strategy … and your data management strategy. Are you sourcing AI agent advice based on, let’s say, support information, but you’re only updating that support material once every six months? That’s not going to work.”

Teams already invested in secure, automated continuous integration and continuous development/deployment pipelines with testing, canary deployment and observability are in the best position to benefit from AI. Organizations risk injecting unvetted AI-generated code without these guardrails into production environments — introducing avoidable security vulnerabilities and compounding tech debt, according to Seroter.

“Now we can do AI-generated code and trust it because we have a good code review process, security scan process, gated releases [and] rollback techniques …  but you’ve got to finish some things first before this is going to be a big benefit to you,” Seroter said.

Equally important is embracing platform engineering to abstract infrastructure complexity for developers. Expecting generalists to be full-stack experts in backend development, security, infrastructure as code and machine learning is unrealistic, according to Seroter. Mature organizations are building internal platforms to deliver scalable services and reusable patterns that let developers focus on delivering business value through AI in application development and beyond.

“There’s still something you’re going to have to have depth in … but, saying ‘I’m going to hire a general-purpose dev who writes front-end, back-end, secures it, talks to databases [and] does [machine learning] …’ you are dreaming,” Seroter added.

Ultimately, successful AI adoption in software development isn’t about skipping steps — it’s about enabling velocity through maturity. Organizations that recognize AI as an augmentation layer, not a replacement for good engineering practices, are best positioned to thrive in this era of accelerated change.

Here’s the complete conversation with theCUBE Research’s Paul Nashawaty and Nathen Harvey, part of theCUBE Research’s AppDevANGLE podcast series:

Image: SiliconANGLE