Agentic AI Is Here — And It's Reshaping How Enterprises Build on the Cloud

For most of the past few years, enterprise AI adoption has been about the model: which LLM to use, how to fine-tune it, where to run it. That conversation is shifting fast. The focus is moving from the model itself to the infrastructure around it — specifically, how AI agents are orchestrated, how they communicate with each other, and how the underlying cloud and data architecture needs to change to support them at scale.

At a recent theCUBE event hosted at the NYSE, industry analysts Rob Strechay and John Furrier of theCUBE Research laid out eight ways agentic frameworks are reshaping enterprise AI strategy. Their analysis is worth paying close attention to — not because it describes what's coming, but because it describes what's already happening in real deployments at major enterprises.

Here's what it means for organizations building or planning AI infrastructure today.

From Isolated AI Tools to Multi-Agent Ecosystems

The first and most significant shift is architectural. Early enterprise AI deployments were largely self-contained: a single model, a single use case, a defined input and output. Agentic frameworks change this entirely. In a multi-agent architecture, individual AI agents — each specialized for a specific task — communicate with one another, pass context between sessions, and coordinate to complete complex, multi-step workflows that no single model could handle alone.

This isn't just a technical detail. It means the entire infrastructure stack — networking, storage, cloud architecture, security, and access controls — needs to be designed with agent-to-agent communication in mind from the start.

MCP: The Protocol Powering Agent Communication

One of the most important developments highlighted in the SiliconAngle analysis is the emergence of Model Context Protocol (MCP) as the de facto standard for how agents communicate with tools, APIs, and data sources. As Strechay described it: "MCP is really the protocol — it's a way to communicate via APIs and basically publish-subscribe."

MCP is gaining rapid adoption across major platforms. AWS Bedrock and SageMaker are both integrating it, and the broader industry is converging on it as the interoperability layer for agentic systems. For organizations building AI infrastructure, MCP compatibility is quickly becoming a baseline requirement rather than an optional feature.

Agent-to-agent (A2A) protocols handle the communication layer between agents in multi-agent systems — and as architectures grow more complex, the design of these communication pathways becomes as important as the models themselves.

Metadata Catalogs as the New Control Plane

Perhaps the most underappreciated infrastructure challenge in agentic AI is metadata management. As agents proliferate across an organization — accessing different tools, data sources, and services — a centralized metadata catalog becomes essential for maintaining visibility and control.

The SiliconAngle analysis identifies metadata catalogs as the emerging control plane for agentic frameworks, centralizing tools, agents, and data across stacks. AWS SageMaker and Bedrock updates reflect this direction. Organizations that don't establish this control plane early will find themselves managing sprawling agent ecosystems with no clear picture of what is accessing what — a governance and security nightmare.

AI-Accelerated Cloud Migration

Agentic frameworks are also changing the economics and timeline of cloud migration. What was historically a multi-year, high-risk infrastructure project is being compressed by AI-infused toolchains that automate discovery, mapping, dependency analysis, and workload portability. Migration that previously took 18 months is being completed in a fraction of the time — with greater accuracy and lower risk.

For organizations still running significant on-premises infrastructure or hybrid environments, this shift matters enormously. The barrier to cloud migration is dropping, and the competitive disadvantage of staying on legacy infrastructure is rising.

Token Economics and Cost-Aware AI Strategy

Running AI agents at enterprise scale is expensive — and the cost structure is different from anything most IT organizations have budgeted for before. Every agent interaction consumes tokens, and in a multi-agent architecture where agents are calling other agents and passing context back and forth, token consumption can scale dramatically.

Intel's Mark Castleman, managing director for Intel AI Cloud, highlighted the rising importance of cost-aware compute strategies as a key enabler of sustainable AI transformation. Organizations need to build token economics into their AI architecture from the start — understanding not just what a model can do, but what it costs to run it at production scale and how to optimize that cost without sacrificing performance.

Security Is Not Optional in Agentic Architecture

This is where AirGap Labs needs to be direct: the shift to agentic AI introduces a fundamentally new security surface that most organizations are not yet equipped to address.

An AI agent with access to your databases, APIs, file systems, and communications tools is a powerful capability — and a significant risk if not properly secured. Prompt injection attacks can hijack agent behavior mid-session. Agents connected to destructive tools can be manipulated into executing actions they were never intended to perform. Context passed between agents can carry sensitive data across boundaries that should be protected. And without a tamper-evident audit trail, there is no way to reconstruct what an agent saw, what it did, or why.

The enterprises succeeding with agentic AI are not the ones moving fastest — they're the ones building security into the architecture from the beginning. That means Zero Trust access controls on every tool and data source agents can reach, inline inspection of agent requests and responses, RBAC policies that restrict what agents are permitted to do, and audit logging that captures every agent decision in a verifiable chain.

What This Means for Your Cloud and AI Strategy

The SiliconAngle analysis makes clear that the vendors and organizations clinging to legacy infrastructure — and the enterprises treating AI security as a bolt-on consideration — risk falling behind as agentic frameworks become the norm.

The organizations pulling ahead are those that have:

• Adopted open protocols like MCP to ensure interoperability across their AI stack
• Built metadata catalogs and control planes that give them visibility across agent deployments
• Designed cost-aware compute strategies that make AI economically sustainable at scale
• Integrated security — including agent-to-agent trust validation, tool RBAC, and audit logging — into their AI architecture from day one

How AirGap Labs Can Help

AirGap Labs sits at the intersection of the infrastructure and security challenges that agentic AI creates. Our cloud practice specializes in designing and deploying multi-cloud architectures across AWS, Azure, OCI, and GCP that are built to support AI workloads — with the networking, security, and data infrastructure that agentic systems require. Our security practice ensures that AI agents operating within your environment do so within properly enforced boundaries, with the visibility and controls to detect and respond when something goes wrong.

And through our partnership with Onnex, we offer AI-Sentinel — an inline security layer that sits between your applications and your AI models, inspecting every agent request and response in real time to block prompt injections, prevent data exfiltration, enforce tool access controls, and maintain a tamper-evident audit log of every agent decision.

Agentic AI is not a future consideration — it's arriving in production deployments now. The infrastructure and security decisions you make today will determine how safely and effectively your organization can harness it.

Contact AirGap Labs at sales@airgaplabs.com or call 949-669-4711 to start a conversation about building AI-ready infrastructure and security for your organization.