Platform Teams Rethink Their Role as AI Reshapes Developer Workflows

Platform engineering teams are being pulled in new directions as AI coding tools become standard fare in developer workflows. During a panel discussion hosted by InfoQ, engineers from major tech companies weighed in on how internal platforms must evolve to support AI-assisted development. The conversation centered on a core tension: how much standardization platforms should enforce versus how much freedom developers need to experiment with AI tools.

Panelists included Stéphane Di Cesare, Davide de Paolis, Stephen Cihak, Camila Macedo, and moderator Renato Losio. They agreed that platform teams can no longer treat AI as an external add-on. Instead, capabilities like model access, prompt management, and evaluation pipelines are becoming core platform services, similar to how CI/CD and infrastructure provisioning were centralized in earlier waves of DevOps.

A key point of debate was the balance between guardrails and autonomy. Too much control over AI tooling can slow down development and frustrate engineers who want to try new models or approaches. But too little oversight risks security vulnerabilities and inconsistent practices. The panelists suggested that platform teams should provide secure defaults and clear policies, while leaving room for teams to customize workflows within those boundaries.

Security was a recurring theme, particularly around code generation and data handling. Developers using AI assistants may inadvertently expose sensitive code or internal APIs to external services. Platforms can mitigate this by offering approved tooling, enforcing data redaction, and integrating security scanning into the AI workflow itself. The panelists also noted that as AI models become more capable, the platform's role in auditing and observability will grow.

The conversation also touched on the shifting skillsets required for platform engineers. Building internal developer platforms now involves understanding model behavior, prompt engineering, and evaluation metrics, not just infrastructure. That means platform teams may need to hire or train for AI-specific expertise, while also educating developers on when to rely on AI suggestions and when to override them.

Looking ahead, the panelists predicted that platform engineering will become even more central as AI adoption deepens. They cautioned against treating AI as a plug-and-play solution, stressing that thoughtful integration requires ongoing iteration and feedback from developers. The open question remains how platforms will keep pace with rapidly evolving AI capabilities without becoming the bottleneck they were meant to eliminate.