Enterprises are integrating artificial intelligence (AI) agents and physical AI into their operations at an accelerating pace, but many are doing so on legacy network foundations that were never designed for machine-speed workloads.
That warning came from Tay Bee Kheng, president of Cisco ASEAN, during the opening address at Cisco Connect 2026 Singapore. She described the current infrastructure as a road network built for bicycles, noting that AI agents place sustained and persistent demand on networks in ways that human-driven traffic never did. “The chatbot has only intermittent pressure and demand on your network. Agents will have a sustained and persistent demand on your infrastructure. Infrastructure that is made for a human click on the network is not applicable for AI agents anymore,” she said.
Why AI agents overwhelm legacy networks
Traditional enterprise networks were designed around human behavior: office workers logging in, browsing applications, streaming video and accessing cloud services. Those patterns are largely predictable, with peak periods and relatively simple east-west traffic flows. AI agents, by contrast, operate continuously, make thousands of micro-requests per minute and need real-time telemetry to make decisions. A single agent may need to pull data from multiple APIs, call a foundation model, update a workflow and return an answer in seconds. When hundreds or thousands of agents are active, the load becomes machine-scale and can no longer be managed with static rules and manual troubleshooting.
Legacy networks also lack the automation needed to keep pace with agent-driven change. Modern AI workloads require dynamic policy updates, automatic bandwidth allocation and programmable infrastructure. Without these capabilities, network teams find themselves constantly reacting to congestion, latency and security incidents. Cisco executives at the event argued that enterprises must move toward secure, self-healing networks that are built with AI-native principles and are able to observe and respond to machine-to-machine traffic.
The shift from human-centric to machine-centric networks also demands a new operational model. Network operations centres that rely on dashboards and manual incident response will not be able to keep up with agentic workloads. AI-powered operations tools can automatically detect abnormal agent behavior, reroute traffic, enforce policy and restore services in seconds. Tay noted that the network itself must become intelligent, not just fast, if it is to support the autonomous workflows that enterprises are beginning to deploy.
A workforce of 10 agents per employee
Beyond the technical limitations of legacy infrastructure, Tay Bee Kheng highlighted the scale of what lies ahead. Organisations must be ready to onboard as many as ten AI agents for every employee, she said, adding a daunting perspective on the AI wave: “There’s no HR system for that at this moment.” Human resources and IT procurement processes were designed for people, not software agents. Enterprises will need catalogs, identity records, permission systems and lifecycle management for agents in the same way they manage employees.
Agentic AI is also changing the nature of work. Instead of tools that wait for someone to type a prompt, agentic systems can reason, plan and act across supply chains, finance operations, customer service and software development. They can coordinate with other agents and make decisions at machine speed. That creates a fundamental security problem: agencies and corporations can no longer assume that every actor on the network is a human employee with a username and password.
Agent identity management will become a core enterprise capability, much like identity and access management for humans. Every AI agent will need a unique identity, an owner, a scope of authority, and a record of its actions. This will be particularly important in regulated industries where audit trails and accountability are mandatory. Without such discipline, enterprises risk creating an invisible layer of automated activity that no one fully understands or controls.
Security blind spots in a machine-driven world
During a media briefing, Robert Pizzari, group vice-president of Asia at Splunk, now a Cisco company, warned about the risks of shadow AI. Employees may deploy unsanctioned foundation models, connect personal AI assistants to corporate data or grant AI agents privileges they should never have. This new form of shadow IT can expose sensitive information, break compliance requirements and create blind spots for security operations teams.
Cisco is tackling the problem with its AI observability stack, which was recently bolstered by the acquisition of Galileo. The stack gives enterprises visibility into model drift and agent behavior, allowing them to detect anomalies before they become incidents. As Pizzari put it, shadow AI will inevitably become a feature of enterprise environments, so organisations also need the ability to hit the handbrake. That means having clear controls, automated responses and a way to terminate a rogue agent on demand.
Zero trust for non-human identities
Koo Juan Huat, Cisco ASEAN’s director of cyber security, stressed that treating AI agents with the same zero-trust architecture used for human employees is the only way forward. He pointed to Singapore’s Government Technology Agency, which recently announced plans to build an AI agent registry for public officers. That registry is an early example of machine identity governance, a practice that will be essential in every sector.
Three actions are required to secure AI agents, according to Koo. First, enterprises must know exactly what agents are running on the network. Second, they must know what each agent is authorised to do. Third, they must implement strict guardrails, such as just-in-time and just-enough permissions. “A human needs to come into the loop and authenticate and authorise the action,” he explained. This keeps humans accountable even when AI systems are executing many processes automatically.
Zero trust also needs to extend to API security. AI agents communicate through application programming interfaces, and those APIs are increasingly the primary attack surface. The legacy mindset of trusting everything inside the network perimeter must be replaced with continuous verification of every request, regardless of whether it comes from a person or a machine.
Fighting frontier AI with frontier AI
As organisations deploy AI, so do threat actors. Rahayu Mahzam, Singapore’s minister of state for digital development and information, reminded the industry that AI-powered voice phishing attacks that cloned CEO voices in 2025 are no longer hypothetical risks. “Agentic AI is here — AI that doesn’t just respond, but reasons, plans and acts,” she said. “But with accelerating capabilities and automation come new digital and cyber risks. AI agents that act without sufficient oversight can cause real harm.”
Cisco is taking an AI-versus-AI approach to defence. Through Anthropic’s Project Glasswing, Cisco used frontier AI models, including Claude Mythos, to scan 1.8 billion lines of code across more than 25 programming languages in eight weeks, achieving a false positive rate of under 3%. This kind of AI-powered code analysis can help enterprises detect vulnerabilities at a scale that human security teams cannot match.
Cisco executives acknowledged that the industry’s security landscape is still too fragmented. To close the gap, Cisco is open-sourcing its AI security and safety frameworks to help the wider ecosystem build secure agentic AI systems. These include DefenseClaw, an open-source framework that scans, sandboxes, and inventories AI agents, their skills and their model context protocol (MCP) connections before they are allowed to run. One component, CodeGuard, performs static analysis on agent-generated code to flag vulnerabilities. These efforts aim to bring the same level of scrutiny to AI agents that traditional security tools bring to applications and devices.
Upskilling people for an AI-led future
Ultimately, the most advanced technology is only as trustworthy as the humans who govern it. At the event, Rahayu Mahzam announced a three-year memorandum of understanding between Cisco and the Digital Defence Alliance Singapore (DDAS) to develop joint training programmes in AI and cyber security. The initiative is designed to level the playing field in AI upskilling and create practical learning opportunities for youths and working professionals.
Already, polytechnic students from the DDAS community are scheduled to visit Cisco’s Tokyo office in October to learn about network security in Japan’s commercial IT industry. The visit follows a similar trip to Seoul in April 2026. Such exchanges are expected to give students hands-on exposure to how large enterprises manage security in real-world environments, while strengthening regional cooperation in AI talent development.
Rahayu concluded with a call to action for the industry. “The question is no longer whether AI will transform the way we work. It already has. The question is whether we are ready to lead that transformation — with skill, with security and with trust at the centre.”
Source: ComputerWeekly.com News