Agentic development demands a multi-model strategy — and the governance to match

The rapid rise of agentic development has radically transformed the software engineering landscape, compelling enterprises to embrace a multi-model AI ecosystem.

The pace of change in software development has outstripped even the most optimistic predictions, leaving enterprises scrambling to keep up. To optimize workflows but avoid provider lock-in, organizations must empower developers to seamlessly orchestrate multiple agents across diverse AI providers, according to Mikhail Vink (pictured), vice president of business development at JetBrains s.r.o., whose integrated development environments serve 15 million developers.

“If you look at AI, this month there is something from Anthropic, next month it’s something from Gemini,” Vink said. “You really, as a developer, need to be top-notch on top of all of those things to get the best from the market.”

Vink spoke with theCUBE’s John Furrier and co-host Alison Kosik at Google Cloud Next, during an exclusive interview on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed the evolving role of the software developer, agentic development and the critical need for robust governance in AI-driven environments. (* Disclosure below.)

The architectural demands of agentic development

While generating code has become commoditized, establishing the infrastructure to manage these complex systems remains a hurdle. Deploying enterprise-grade solutions requires more than casual experimentation, Vink noted. The reality is a sprawling web of interconnected components — agents, memory layers, data pipelines and external tool integrations — all of which must be configured and maintained continuously.

“You need to have a bunch of agents … you need to control them; but then the story is that you also need to pass the data, pass the context, pass the memory,” Vink said. “You need to connect them to the [Model Context Protocol server] to get the context to the agents, because otherwise, you’re going to be missing out on the real-world data — structured data. You need to configure a lot of things for your developer environment to be sustainable at this point.”

In response, JetBrains has been building a governance platform designed to track costs, monitor model access and analyze what developers are actually accepting from AI-generated suggestions — giving enterprises the visibility they need to manage AI at scale. This requires developers to act as orchestrators, elevating the importance of quality assurance and critical thinking. Engineers must dive deep into generated algorithms to ensure security and functionality, Vink noted.

“I would say that the most critical thing [in quality assurance] is critical thinking,” he said. “It’s not just approving what AI gives you, what agents generate — it’s going really deep and understanding how the system works.”

Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of Google Cloud Next:

(* Disclosure: JetBrains sponsored this segment of theCUBE. Neither JetBrains nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Photo: SiliconANGLE