The Year-Three Reckoning: Agent Deployments Have Outpaced the Workforce That Runs Them

Posted

CompTIA has announced plans to develop CompTIA AgentOps+, a vendor-neutral credential designed for professionals tasked with deploying, securing, monitoring, and governing AI agents. The move, detailed in a press release dated Oct. 23 on CompTIA’s own newsroom and syndicated a day earlier via PR Newswire, marks a change in what the industry is measuring: not what agents can do, but who is accountable for them.

As Carl Bowman, senior vice president of exam services at CompTIA, stated: “The focus has been on what AI agents can do, but the more urgent question is who will secure, govern and be accountable for them. As agents gain access to enterprise data and workflows, organizations need professionals who can manage them securely and responsibly. CompTIA intends to help define that standard.”

This credential is an admission that the current workforce is not yet equipped for the scale of agent deployments already underway. While the certification is planned for release in Q2 2027, its existence highlights a structural gap in the market. Enterprises are currently rushing to deploy agents, yet the specialized labor required to maintain them remains largely undefined and uncertified.

The scale of these deployments is significant. Uber’s rollout of ChatGPT Work to more than 30,000 employees – up from a 1,000-employee pilot, at an undisclosed cost, with the company measuring adoption and outcomes rather than productivity gains – sets the 30,000-seat benchmark other deployments get measured against. Cisco is rolling out its MyAgent tool to 90,000 employees on its Circuit platform, according to a corporate announcement. The 90,000 figure comes from that announcement; Cisco’s FY2026 Form 10-K lists a total headcount of 82,400. Meanwhile, vendor-reported data from the Microsoft 2026 Work Trend Index indicates active agents in the Microsoft 365 Copilot Agents platform and SharePoint grew 15x year over year, and 18x in large enterprises – with no published baseline or denominator.

Despite this rapid adoption, the infrastructure for agent governance is lagging. A KPMG survey cited inside CompTIA’s release puts 54% of organizations actively deploying AI agents, while a Deloitte consultancy survey cited in the same release finds only 21% of companies have a mature governance model for autonomous systems. The skills side reads the same way: a secondhand compilation of DataCamp/YouGov survey data shows 82% of enterprise leaders say their organization provides AI training, yet 59% still report an AI skills gap and only 35% report a mature, organization-wide upskilling program. Deloitte’s enterprise survey, also carried secondhand, ranks insufficient worker skills as the top obstacle to integrating AI into existing workflows.

The Cost of Control

Enterprises are paying for the agent seats, but the operating labor behind them is treated as an unpriced variable. Our reporting on the unpriced reviewer established that human review is load-bearing and priced at zero, and it is routinely left out of the initial cost-benefit analysis. The result shows up in our coverage of enterprises buying four layers of control before proving one layer of value.

The market already shows the friction. Secondhand market research from S&P Global Market Intelligence and Forrester puts roughly 11% of enterprises running an agent in production at scale, with about 88% of agent pilots never reaching production. A secondhand Gartner forecast goes further: more than 40% of agentic AI projects could be canceled by the end of 2027 on cost, unclear value and weak risk controls.

CompTIA’s planned credential covers deploying and integrating agents, securing identities and data access, monitoring performance and identifying risks, and it is aimed at DevOps, cloud operations, platform engineering, systems administration and cybersecurity staff. What it does not cover yet is any actual test: no exam code, price, beta window or objectives have been disclosed. Until the workforce catches up to the deployment scale, the burden of governance falls on teams already running agents with no standard to run them against.

Agents at Work, money