The Evolution of Agentic AI and its Governance Challenges

Agentic artificial intelligence is experiencing significant acceleration, particularly in the Asia-Pacific (APAC) region. This evolution, however, introduces new complexities and risks for enterprises. Without unified visibility and control over agentic AI systems, security gaps emerge, costs can spiral uncontrollably, and compliance verification becomes extremely difficult. Kong Inc., a leading developer of API and AI connectivity technologies, highlights how enterprises in APAC, as they move AI from experimentation to production, face growing connectivity and governance challenges.

Modern organizations do not simply route prompts to a single LLM. Instead, they orchestrate complex systems where agents communicate with external tools via protocols like MCP (Multi-Cloud Protocol) and collaborate with other agents using emerging protocols, such as Agent-to-Agent (A2A). Most current solutions address only a fraction of this traffic, forcing enterprises to stitch together multiple point solutions or accept blind spots in their AI infrastructure.

Kong Agent Gateway: A Unified Control Point

Alex Drag, head of product marketing at Kong, emphasized how the landscape has radically changed. Initially, LLM governance was relatively contained: a request went to a model, a response came back, and an AI Gateway mediated to enforce policy. Today, agentic architectures are far more complex: agents call tools via MCP, delegate tasks to other agents via A2A, and these agents produce and consume event streams. Data flows in every direction, often with little to no human intervention, and with poor visibility into what is actually happening.

To address this complexity, Kong recently released Kong Agent Gateway, part of the AI Gateway 3.14 release. This solution was designed to be the only gateway on the market to support advanced use cases including LLMs, MCP, and A2A communications. With Agent Gateway, organizations gain a definitive control point for the entire AI lifecycle, centralizing visibility and policy enforcement across the full AI data path, from APIs and LLMs through to MCP tools and agent-to-agent workflows.

Transition to Commercial Automation and Data Sovereignty

Ruiguo Lai, regional sales director Asia at Kong, highlighted how for APAC organizations, the transition from experimental AI to production-grade autonomy represents a crucial change in regional competitiveness. As enterprises in ASEAN, South Korea, and the Greater China Region scale multi-agent systems in regulated and multi-cloud environments, the primary challenge shifts from simple model access to the complex governance of “Agent Sprawl.” This involves managing operational risk and unpredictable token costs, securing the “Agent Mesh,” and ensuring that autonomous A2A communications remain within the guardrails of regional data sovereignty laws and internal security protocols.

The introduction of Kong Agent Gateway enables a seamless flow of intelligence without sacrificing governance. This is particularly vital in the APAC region, where the need for rapid digital commerce innovation must coexist with stringent compliance requirements in financial services and government sectors. For those evaluating on-premise or hybrid deployments, solutions like Agent Gateway offer analytical frameworks to assess trade-offs between control, security, and TCO, which are fundamental aspects for data sovereignty and compliance.

Concrete Benefits for Enterprises

With Agent Gateway, organizations can achieve a range of tangible benefits. These include unified observability across all native AI traffic, with Kong Konnect acting as a central dashboard. The solution enables production-ready agentic AI, offering essential security features, access control, and audit capabilities to confidently move agentic workloads from pilot to production. Furthermore, it provides cost visibility and control, thanks to granular tracking of token consumption and resource usage in agent workflows, allowing for accurate cost allocation and margin optimization.

Reza Shafii, SVP of Product at Kong, reiterated that every enterprise grapples with the same three challenges: not having full visibility into all AI traffic and resource consumption in an agentic workflow, struggling to adopt AI in a way that helps increase margins, and dealing with issues when moving AI and agentic workloads into production. Agent Gateway was developed precisely to solve these problems, enabling engineering teams to govern all their multi-agent traffic in a single place, providing the control and connectivity necessary to make agentic AI workable at enterprise scale.