CompTIA announced plans for CompTIA AgentOps+, a credential "intended to establish a vendor-neutral standard for professionals who deploy, secure, monitor, and govern AI agents." The announcement opens with a sentence most engineering leaders will recognize: AI agents are moving into enterprise operations, "but the skills needed to manage them responsibly remain undefined."
The credential and its learning products are planned for release in Q2 2027. Meanwhile, KPMG's Q3 2026 AI Quarterly Pulse, published September 24, found that 62% of organizations are building, deploying, or developing AI agents, up from 53% the previous quarter. Multi-agent systems climbed to 25% from 6%. The agents are arriving well before the people certified to run them.
The same KPMG survey found a warning sign. The share of organizations building controls into their agents, along with monitoring and evaluation procedures, fell to 30%, down from 43% two quarters earlier. Deloitte's State of AI in the Enterprise 2026 found that only 21% of respondents say their organizations have a mature governance model in place for agentic AI.
What Is AgentOps, and Why Does It Need Its Own Role?
AgentOps is the work of running AI agents in production: deploying and integrating them, securing their identities and access, monitoring their behavior, responding to incidents, and keeping humans accountable for what they do. CompTIA's planned credential covers those five areas. It is closer to DevOps and security operations than to data science.
CompTIA lists the five skill domains as 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; and supporting governance, accountability, auditability, and human oversight. That list reads less like an AI research job and more like an operations job.
The audience confirms it. CompTIA says the credential is intended for professionals in DevOps, cloud operations, platform engineering, systems administration and cybersecurity. Carl Bowman, CompTIA's senior vice president of exam services, framed the gap this way: "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."
In our recruiting work, most companies have not named this role yet. Agent operations usually split between the engineers who built the agent, the platform team that hosts it, and a security team that learns about it late. That works for a pilot. It stops working when agents touch customer data, financial systems, or production code every day.
Why Can't You Wait for the Certification?
Because your agents will not wait. KPMG's Q1 2026 pulse found that 57% of leaders expect people to manage and direct AI agents, and 63% now require human validation of agent outputs, up from 22% in Q1 2025. Someone on your team is already doing that validation, whether or not it appears in their job description.
A certification also takes time to mean something after it launches. Training providers need courses, candidates need study time, and hiring managers need to learn what the credential does and does not prove. Even if AgentOps+ ships on schedule, it is reasonable to expect most of 2027 to pass before certified candidates are common. That estimate is ours, not CompTIA's.
There is also no reliable count of the role yet. We looked for independent 2026 data on job postings for AgentOps, agent operations, or agent reliability roles and found none. Lightcast counted more than 86,000 job postings mentioning agentic AI skills in 2025, but it does not break out operations roles. You are hiring for a job the labor market has not labeled.
Who on Your Team Could Run AI Agents Today?
Start with the people CompTIA is targeting. Engineers in DevOps, site reliability, cloud operations, platform engineering, and security already own deployment, monitoring, access control, and incident response for your other production systems. In our experience, adding agents to that scope is a shorter leap than teaching agent operations to someone who has only built models.
Site reliability and DevOps engineers bring monitoring, alerting, on-call habits, and incident reviews. They usually need to learn how agents fail differently: loops, unexpected tool calls, drift after a model update, and outputs that look correct but are wrong. Those failure modes need new signals, not just new dashboards.
Security and identity engineers bring least-privilege access, secrets management, and audit trails. Agents act through credentials, and each agent needs its own identity, scoped permissions, and a clear owner. Our guide to agent governance covers the policy side, and the security engineer is often the right person to enforce it day to day.
Platform engineers bring the shared tooling: gateways, deployment pipelines, cost controls, and the internal services agents call. In multi-agent systems, controls belong at the orchestration layer, which is platform territory. Our guide to multi-agent AI controls explains why that layer matters as adoption grows.
How Do You Vet AgentOps Skills Without a Credential?
Test the work, not the résumé. With no credential to lean on, the most reliable signal is how a candidate handles a realistic agent problem. These five practices map to CompTIA's planned domains, so the hires you make now should still look right once the certification exists.
Run a deployment exercise. Give candidates a simple agent and ask how they would ship it: environments, rollback, versioning of prompts and models, and what they would check before go-live. Strong candidates ask what the agent can touch before they ask which framework it uses. Weak ones focus on the model and ignore the integration.
Ask them to scope an agent's access. Describe an agent that reads tickets, queries a customer database, and sends emails, and ask them to design its permissions. Listen for one identity per agent, least privilege, separate credentials from humans, and a named owner. This tests the identity and access domain directly and reveals security instincts quickly.
Hand them a failing agent. Share logs from an agent that looped, called the wrong tool, or produced a plausible but wrong answer, and ask them to diagnose it. Good AgentOps candidates look for traces, step counts, cost per task, and changes in the underlying model, not just error messages. In our recruiting work, this is the skill we see missing most often.
Walk through an agent incident. Ask how they would respond if an agent sent incorrect information to customers for two hours. You want containment first, then a clear record of what the agent did, who approved it, and what changes next. In our experience, candidates with real incident experience in other systems adapt their playbook well.
Check governance in practice. Ask how they would prove to an auditor what an agent did last Tuesday and who was accountable. Look for logging, approval points, and documented human oversight. Our guide to vetting remote engineers covers identity verification for the hiring process itself.
Should You Hire, Train or Borrow AgentOps Talent?
Usually all three, in sequence. In our experience, the fastest path is to name an internal owner from SRE, platform, or security, fund their training now, and bring in outside capacity for the first production rollouts. Our 2027 AI skills plan guide covers when to build, buy, or borrow.
If a vendor built your agents, make AgentOps part of the handover. Ask who will monitor, patch, and govern the agent after the vendor's engineers leave, and get the runbooks, access maps, and logs in writing. Our guide to who owns a vendor-built AI agent covers that transition in detail.
Common Questions About AgentOps
What is the CompTIA AgentOps+ certification?
CompTIA AgentOps+ is a planned vendor-neutral credential for professionals who deploy, secure, monitor, and govern AI agents. CompTIA announced it on October 22, 2026, and plans to release it with related learning products in Q2 2027. It targets DevOps, cloud operations, platform engineering, systems administration, and cybersecurity professionals.
What does an AgentOps engineer do?
An AgentOps engineer runs AI agents in production. The work covers deploying and integrating agents, securing their identities and permissions, monitoring their behavior and health, responding to agent incidents, and supporting governance and human oversight. It is an operations role, closer to site reliability and security engineering than to model development.
Who should own AI agents in production?
Name a single accountable owner for each agent in production, usually from site reliability, platform engineering, or security. KPMG's Q3 2026 pulse found that the share of organizations building controls into their agents alongside monitoring fell to 30%, suggesting many agents run without clear operational ownership. Builders shouldn't be the only ones watching them.
How do you hire for AgentOps before the certification exists?
Recruit from DevOps, SRE, platform, and security teams, then test practical skills. Use exercises on deployment, access scoping, debugging a failing agent, incident response, and audit evidence. These map to CompTIA's planned skill domains, so the people you hire now should still meet the standard once the credential is available.
If you are putting agents into production and nobody owns running them yet, talk to the Golabs team. We can help you define the role, vet candidates against real agent scenarios, or provide dedicated AI teams and AI talent while you build the capability in-house.