Gartner Predicts 70% of Enterprises Will Abandon Vendor-Built Agentic AI by 2028, Shifting Platform Engineering Focus

Gartner has released a significant prediction, stating that by 2028, 70% of enterprises will abandon agentic AI solutions built by vendor-forwarded engineering. This forecast suggests a growing disillusionment with off-the-shelf or heavily vendor-dependent agentic AI implementations, pushing organizations towards more internally developed and controlled solutions. The announcement, made today, points to a strategic re-evaluation of how enterprises integrate and manage AI, particularly agentic systems, within their operational frameworks. This prediction is highly significant for platform engineering teams and practitioners. It signals a shift from simply consuming vendor-provided AI services to actively building and owning the underlying platforms that enable agentic AI. For platform engineers, this translates into a heightened demand for skills in AI infrastructure, MLOps, and the development of internal developer platforms (IDPs) that can effectively support autonomous AI agents. The implication is that organizations are realizing the need for greater control over their AI deployments, especially concerning data governance, security, and the ability to customize AI behavior to specific business logic. Relying on external vendors for core agentic AI capabilities can lead to integration challenges, lack of transparency, and difficulties in adapting to evolving business requirements or regulatory landscapes. This trend aligns with the broader movement in cloud-native and DevOps towards internal developer platforms and self-service capabilities. For years, the industry has emphasized empowering development teams with the tools and infrastructure to operate independently, often through well-defined "golden paths" and self-service portals. The integration of AI, particularly agentic AI, is now extending this paradigm. Platform engineering is evolving to encompass AI-native infrastructure, where AI agents are treated as first-class users of the platform, complete with their own access controls, resource quotas, and governance policies. The focus is on creating a secure, observable, and cost-effective environment for AI workloads, moving beyond traditional application delivery to include model versioning, inference optimization, and the operational patterns unique to AI. This evolution is also driven by the increasing complexity and resource intensity of AI workloads, making efficient internal management crucial. In practice, this means platform engineering teams should prioritize developing capabilities that enable the secure and governed deployment of agentic AI. This includes building robust internal platforms that offer self-service provisioning for AI-specific resources (like GPUs), establishing clear governance frameworks for AI agent behavior, and implementing FinOps practices tailored to AI workloads to manage token and inference costs effectively. Practitioners should focus on creating "agent golden paths" that guide the development and deployment of AI agents, ensuring they adhere to organizational standards for security, compliance, and cost. Furthermore, platform teams will need to develop expertise in areas like AI agent orchestration, model registries, and AI-aware observability to provide the necessary support for internal AI initiatives. The trade-off for greater control and customization is the increased responsibility for building and maintaining these complex AI platforms internally, requiring significant investment in talent and infrastructure. Ignoring this shift risks falling behind organizations that are strategically investing in their internal AI platform capabilities.