What Lies Behind Agentic OPS: Security and Orchestration

Is your company ready for an AI agent to authorize, on its own, a wire transfer, a purchase or a critical configuration change?

The trend: from AI that assists to Agentic OPS, AI that executes

What matters today is where the enterprise AI conversation has moved: from a model that answers questions to multi-agent systems, sets of autonomous agents that interact, exchange information and make decisions based on internal generative models, without a person approving every step (Tian et al., 2025). This is what is now known as Agentic OPS, or agentic operations: business processes orchestrated by AI agents that both recommend and execute.

The pace of adoption supports the trend. Gartner estimates that by the end of 2026, 40% of enterprise applications will have embedded AI agents for specific tasks, a leap from less than 5% in 2025 (Gartner, 2025). The study by Tian et al. (2025) also cautions that these multi-agent systems do not automatically become more robust as more agents are added: their real value depends on how they coordinate, and so does their main source of risk.

The application: when agents talk to each other, who secures the conversation?

The convenience of using agents is already settled; what remains open is how prepared the infrastructure is to support them securely. If different AI systems are going to communicate with each other to execute business-critical tasks, such as approving a payment, updating inventory or escalating a ticket to another system, that communication must be secured. Consider a conversational agent manipulated through a malicious instruction hidden in an email or a document into authorizing a fraudulent purchase or disclosing sensitive information.

This scenario already has a technical name. The OWASP Top 10 for LLM Applications identifies Excessive Agency (LLM06:2025) as one of the ten most critical risks: a model or agent with more capability, autonomy or permissions than its task actually requires (OWASP Foundation, 2025). In a single-agent setup, that risk is serious. In Agentic OPS, where multiple agents pass information and decisions to one another, the problem multiplies: the dynamics between agents can amplify an isolated vulnerability instead of containing it (Tian et al., 2025).

In business terms, this translates into a very concrete question for any CIO or CISO: if a procurement agent communicates with a finance agent to close a transaction, who audits that conversation, who segments it, and what happens if either end has been compromised?

The decision: secure orchestration starts at the network

This is where Agentic OPS stops being an issue exclusive to the data team and becomes an infrastructure issue. Model-level controls (restrictive prompts, permission policies, output validation) are necessary, but they are not enough on their own: they cannot stop a compromised agent from moving laterally to another system if the network allows it. Governing autonomous agents needs to rest on the same discipline already applied to any identity that connects to the corporate network: segmentation, least-privilege access control and visibility into every communication, now extended to software agents as well as people.

This is, precisely, the digital highway behind Agentic OPS: the network infrastructure that determines whether communication between agents is visible, segmented and auditable, or whether it happens in a blind spot. At Ikusi, through our Enterprise Networking solutions, we help companies build that foundation: secure architectures with end-to-end visibility, capable of supporting both human user traffic and a new generation of agents making business decisions in real time.

Is your network ready to govern the next generation of AI agents? Let’s talk about how Ikusi Enterprise Networking can give your Agentic OPS operation the security and orchestration foundation it needs.

References

Gartner, Inc. (2025, August 26). Gartner predicts 40% of enterprise apps will feature task-specific AI agents by 2026, up from less than 5% in 2025 [Press release]. https://www.gartner.com/en/newsroom/press-releases/2025-08-26-gartner-predicts-40-percent-of-enterprise-apps-will-feature-task-specific-ai-agents-by-2026-up-from-less-than-5-percent-in-2025

Ikusi. (n.d.). Redes empresariales [Enterprise networking]. Enterprise networks – Ikusi

OWASP Foundation. (2025). OWASP Top 10 for large language model applications. https://owasp.org/www-project-top-10-for-large-language-model-applications/

Tian, F., Luo, A., Du, J., Xian, X., Specht, R., Wang, G., Bi, X., Zhou, J., Kundu, A., Srinivasa, J., Fleming, C., Zhang, R., Liu, Z., Hong, M., & Ding, J. (2025). An outlook on the opportunities and challenges of multi-agent AI systems. arXiv. https://arxiv.org/abs/2505.18397