AI Agents Are Becoming a Workforce Category

For most of the past decade, AI in customer service was classified as software. Companies deployed chatbots, virtual assistants, and automation tools to reduce workloads and deflect contacts. Success was measured in containment rates and cost savings. This is changing.

A series of recent announcements from Cisco, Genesys, and NiCE reveal a shift in how leading CX vendors are conceptualising AI in the contact centre. Rather than positioning AI as a tool that supports human work, vendors are increasingly treating AI agents as a workforce category in their own right, which must be deployed, monitored, managed, and evaluated much like human employees.

Traditional automation was deterministic. It performed predefined tasks along fixed logic paths, and its behaviour was predictable and auditable. Now, modern AI agents hold conversations, complete multi-step workflows, make recommendations, take actions on behalf of customers, and can adapt their behaviour based on context, memory, and available information.

This has operational implications. When AI begins performing work rather than simply supporting work, organisations need new frameworks to manage it. The question is no longer only how to deploy AI, but how to supervise, evaluate, and improve it over time. The industry is gradually moving from automation management to what might be described as digital labour management.

The scale of that challenge may be larger than many organisations currently anticipate. Gartner predicts that by 2028 the average Fortune 500 enterprise could have more than 150,000 AI agents in operation, creating significant governance, accountability, and lifecycle management requirements as agent populations expand. As AI adoption grows, managing digital labour may increasingly resemble workforce management as much as software administration.

One of the clearest demonstrations of this shift has come out of Cisco Live 2026, where the company unveiled a suite of products it grouped under the concept of the "agentic workforce". The launch included AI Agent 360, a monitoring and observability tool for AI agents, and what Cisco refers to as a complete rebuild of workforce tooling for environments where human and AI agents coexist.

Vinod Muthukrishnan, Vice President and General Manager of Webex CX at Cisco, writing on the Cisco blog ahead of the announcements, framed the operational challenge directly: "Managing a workforce of agents that includes both human and AI is a new operational challenge, and most WEM tools simply weren't built for it."

The rebuilt Cisco AI Workforce Engagement Management suite covers forecasting, scheduling, quality management, performance dashboards, real-time agent guidance, and onboarding. Critically, every capability is designed to work across both human and AI agents, a deliberate departure from legacy WEM tools built when every agent was human and every evaluation was manual and sampled.

Genesys has arrived at a similar conclusion through a governance-first approach, applying controls and oversight mechanisms more commonly associated with managing a human workforce. The company has embedded AI Guides directly into its Genesys Cloud platform, providing defined guardrails, permissions, and transparency controls for agentic AI behaviour. The rationale is partly commercial, partly regulatory. A Genesys survey found that over a third of CX leaders lack formal AI governance policies, creating potential risks as AI systems become more autonomous and take on increasingly complex responsibilities.

The governance challenge becomes more significant when AI agents are viewed as workers rather than tools. Different agents perform different tasks, carry different risks, and increasingly require different forms of oversight. Gartner has argued that different types of AI agents may require different governance approaches depending on their role, autonomy, and level of risk, suggesting that oversight of AI workforces could become considerably more complex as adoption grows.

One of the defining risks of agentic AI is invisibility. Unlike traditional software, AI agents can behave differently from one interaction to the next, learn from context, and produce outputs that are difficult to predict in advance. This has created growing demand for AI observability: the ability to monitor, audit, and understand what AI agents are actually doing.

Cisco's AI Agent 360 is designed specifically to address this. The product provides monitoring, security, and observability capabilities for AI agents, giving contact centre supervisors visibility into agent behaviour, performance, and escalation patterns. The positioning mirrors what quality monitoring has historically done for human agents, applied to a population that operates at far greater scale and speed.

NiCE has similarly positioned its CXone platform around managing human and AI agents as part of a unified workforce, embedding AI throughout workforce engagement management and workforce intelligence capabilities.

Managing AI agents as a workforce category also requires new performance metrics. Traditional workforce management focuses on adherence, occupancy, and average handling time. These measures were designed for human workers, and they map poorly onto AI agents whose capacity is theoretically unlimited and whose "off-task" behaviour looks nothing like a human going on break.

Contact centre leaders managing hybrid workforces may increasingly need to evaluate AI agents on resolution quality, escalation rates, decision accuracy, customer trust scores, and the frequency of human intervention. These are not simply adjusted versions of existing metrics; they reflect a fundamentally different kind of operational accountability. Cisco's AI Agent 360 and NiCE's AI-powered workforce intelligence capabilities both reflect this shift, focusing less on traditional workforce efficiency metrics and more on monitoring outcomes, behaviours, and intervention patterns.

The industry's debate about AI often centres on replacement. A more accurate description of what is currently taking shape is coexistence. AI agents are handling routine, high-volume interactions at scale, while human agents are focusing on complex, emotionally nuanced, and high-stakes conversations. Designing workflows that manage both populations effectively has become one of the defining operational challenges in CX, and neither population currently has frameworks fully adequate to the task.

This distinction may sound semantic, but it has practical consequences. Software is purchased and deployed. Workforces are recruited, trained, monitored, governed, and evaluated. As organisations begin applying those same disciplines to AI agents, the management of digital labour may become as important as the deployment of AI itself.