A cybersecurity professional studies a dashboard connecting AI assistants, cloud services, and security controls.
CompTIA says it is building a certification for the people who run AI agents, and it isn't available yet. The company announced AgentOps+ on October 22, 2026. The credential is aimed at sysadmins, cloud and DevOps engineers, platform teams and security staff. Those are the people who already handle Entra identities, Intune policies, Defender alerts and Copilot rollouts in Microsoft shops.

The story deserves some care, because the headline claims more than the evidence shows. This piece covers what CompTIA actually announced, what it left out, and how to read the adoption and governance statistics attached to it.

A cybersecurity professional studies a dashboard connecting AI assistants, cloud services, and security controls. What CompTIA actually announced​

AgentOps+ is meant to be a vendor-neutral standard for professionals who deploy, secure, monitor and govern AI agents, spanning AI, cybersecurity, cloud infrastructure and IT operations. The certification and a suite of related learning products are planned for release in Q2 2027. Today, nobody can sit this exam.

The planned skill areas are:

  • Deploying and integrating AI agents
  • Securing agent identities, permissions and data access
  • Monitoring performance, behavior and operational health
  • Identifying and responding to agent-related risks and incidents
  • Supporting governance, accountability, auditability and human oversight

CompTIA says it is designing the credential around durable operational principles rather than the products of a single provider. That matters for Windows and Microsoft 365 admins. An agent built in Copilot Studio, one running on Azure, and one bought from a third party all share the same underlying problems: identity, permissions, logging and kill switches.

What is still missing​

CompTIA's release gives no exam code, price, beta window or detailed objectives. Third-party coverage fills in some of that, but it is shaky. One certification-comparison site says a draft with the code AG0-001 exists, with four domains: architecture, deployment and configuration; governance, security and compliance; observability, performance and cost management; and agent operations. The same site says draft objectives change before launch, so wait for the final version before buying study material. I couldn't confirm the draft from CompTIA's own announcement, so treat it as a lead and not as the exam blueprint.

Admins who want to start now have a few existing options:

  • CompTIA's current AI security exam, SecAI+, which one comparison site says launched on February 17, 2026.
  • CompTIA's AI Agent Essentials course, described as a 4–5 hour, hands-on course for non-technical and lightly technical professionals. It targets a different audience from AgentOps+.

Reading the numbers without fooling yourself​

The CompTIA release cites two surveys, and the coverage that followed set them side by side as if they matched. They don't.

KPMG, 54% deploying agents. The release cites KPMG's first-quarter 2026 AI Quarterly Pulse Survey, which found more than half of organizations are actively deploying AI agents. KPMG's own page describes a narrow sample: the US survey covers 237 C-suite and business leaders at organizations with at least $1 billion in annual revenue. That is not a census of the economy. "Actively deploying" also doesn't mean agents are running in production at scale.

Deloitte, 21% with mature governance. Deloitte's survey covered 3,235 IT and business leaders in 24 countries. Only 21% said their organizations have a mature governance model for agentic AI. This is a self-reported maturity measure and not an audit. Deloitte's description of mature governance includes clear limits on what agents can decide alone, real-time monitoring that flags anomalies, and audit trails of agent actions. That is a useful checklist for any admin.

The two surveys use different populations and different questions. They suggest a gap between deployment and governance. They don't prove that the other 79% have no controls.

KPMG's own findings add some texture. Its report says 63% of leaders now require human validation of AI agent outputs, up from 22% in Q1 2025, and that skills gaps are a leading barrier. That is evidence organizations are adding oversight, but it comes from the same kind of survey.

The Microsoft telemetry, in context​

The original story leans on Microsoft's 2026 Work Trend Index, which reports 15x year-over-year growth in active agents, and 18x in large enterprises. Microsoft's methodology says this counts unique active agents across the Microsoft 365 Copilot Agents platform and SharePoint agents, with an agent counted as active if it had at least one day of user-initiated use or one autonomous run in a rolling 28-day period. The data covers March 2025 to March 2026.

There are three limits worth knowing:

  • It is a growth multiple. Microsoft's own chart notes say all metrics are expressed as shares and ratios, with no absolute counts. A 15x rise from a tiny base is still a small number.
  • It measures Microsoft's own ecosystem, so it says nothing about industry-wide totals.
  • Low-code agents built by employees count too. That is exactly where an admin's governance headaches come from, and it makes the figure relevant to anyone running a Microsoft 365 tenant.

Uber and Cisco: scale is not outcome​

OpenAI's announcement says Uber is rolling out ChatGPT Work to more than 30,000 employees, expanding from an initial pilot of 1,000. The rollout will connect to tools such as Google Drive, Gmail, Google Calendar and Slack. It will be backed by an internal AI skills library, approved integrations and employee training. OpenAI says the companies plan to share insights on adoption and measurable outcomes as the rollout proceeds, so no productivity results have been reported yet.

The original story also says Cisco is rolling out a MyAgent tool to 90,000 employees. I could not corroborate that figure from an official Cisco announcement, so I'm not repeating it as fact.

Even a verified seat count shows access. It doesn't show how many agents run autonomously, what they can touch, or how many people it takes to keep them in check.

What the evidence does not support​

The headline's framing says millions of agent deployments already depend on a missing certified workforce. The evidence doesn't show that. The original story also repeats claims about training gaps, pilot failure rates and Gartner's project-cancellation forecast. I couldn't check those against original research, so I've left them out. The Gartner figure in particular is a forecast and not an observed result.

CompTIA also has a commercial interest. A certification body saying the skills are "undefined" is making the case for its own product. That doesn't make the point wrong, since Deloitte's governance finding is real. But no evidence shows certification is needed to operate agents safely. Good controls come from technical practice and policy.

What admins can do now​

Don't wait until Q2 2027. The five AgentOps+ topic areas already work as a rough audit plan for any tenant running agents:

  1. Inventory agents. List which agents exist, who built them, and what data they can reach.
  2. Treat agents as identities. Give each one scoped permissions, an owner, and a review date.
  3. Turn on logging. Make sure agent actions land in an audit trail you can search.
  4. Define human approval points. Decide which actions an agent can take alone and which need sign-off, as Deloitte's governance description suggests.
  5. Write an incident playbook. Know how to suspend an agent and revoke its access quickly.

The takeaway​

CompTIA's announcement is a credible signal that agent operations is becoming a recognized IT job. It is not proof of a workforce crisis. The strongest figures behind the story are survey results and growth multiples from vendors and consultancies, each with its own limits. For now, the useful thing about AgentOps+ is its outline of what agent operations involves. The exam itself, with its price, objectives and date, is still to come.

 

References

  1. The Year-Three Reckoning: Agent Deployments Have Outpaced the Workforce That Runs Them - Forkast News Forkast News 2026-10-24T14:43:23+00:00
  2. Uber brings ChatGPT Work to all employees with OpenAI openai.com
  3. CompTIA launches effort to define workforce skills for AI agents comptia.org