Most SOC Leaders Redesigning L1 Roles as AI Agent Adoption Grows
Security operations teams are changing how entry-level analysts work as AI agents become more common in security operations centers (SOCs), but organizations are more often redeploying or retraining staff than eliminating the roles. A Strike48 survey found that 87% of security leaders are redesigning their SOCs in response to AI agents.
The findings come from Strike48’s report, The State of Agentic Security H2 2026, based on responses from 534 security and IT leaders. Respondents worked at organizations with at least 200 employees and a security operations function in the UK, Canada and the US. The survey indicates a sharp rise in reported deployments: 74% said their SOCs had deployed agents, compared with 36% in May. Nearly half, 49%, said they were using agents across multiple use cases.
Entry-level roles are being reshaped
Rather than reporting broad replacement of Tier-1 analysts, respondents described several ways they are adapting roles and skills. Almost half, 49%, said they were hiring for capabilities such as AI detection engineering and automation. Another 42% said they were retraining L1 analysts to supervise agents, while 38% were moving those analysts into investigation work.
The survey also found that 30% of respondents were expanding their teams by hiring more entry-level analysts. Taken together, these results suggest that some organizations expect human staff to remain part of agent-supported operations, with duties shifting toward oversight, investigation and automation-related work. The figures describe respondents’ reported plans and practices; they do not establish how roles are changing across the industry as a whole.
The discussion matters because entry-level SOC work has traditionally offered analysts opportunities to develop experience. A separate Swimlane report cited in the survey coverage found that 24% of SOC analysts felt AI had limited their ability to build skills, and 10% expected fewer opportunities for junior analysts to gain experience. Those concerns underline a practical workforce challenge: organizations adopting agents need to consider how junior employees will continue to learn while routine tasks are automated or redistributed.
Trust remains a limit on autonomy
Deployment does not necessarily mean that agents are permitted to act independently. Only 30% of Strike48 survey respondents said their teams allowed agents high or full autonomy. Nearly half, 47%, said humans execute actions or approve them in advance. A further 23% said agents could act within defined boundaries.
Respondents identified several conditions they considered important for trusting agents with more autonomous work. The leading requirement was demonstrated accuracy in the organization’s own environment over time, cited by 57%. A complete audit trail of actions and the data supporting them came next, at 53%. Technical controls that prevent an agent from exceeding its limits were named by 46%, while 44% emphasized defining a precise scope for each agent. Reliable rollback or undo capabilities were important to 40%.
These priorities point to operational controls that matter when agents are connected to SOC workflows. Accuracy needs to be assessed against the organization’s alerts and environment, while audit records can help teams review what an agent did and why. Enforced boundaries and a defined scope are distinct from relying on instructions alone: the survey respondents specifically highlighted technical limits. Rollback also addresses the possibility that an authorized action may need to be reversed.
Workflow design and oversight
Strike48 co-founder and president Tim Leehealy said that some teams begin with a single agent and a broad instruction to perform Level 1 analysis, then connect it to operational tools. He argued that this approach can fail to capture the full triage process, and that teams progressing toward greater autonomy divide work into narrower tasks and deterministic workflow steps. This is his assessment of deployment practices, rather than a reported comparative test of different designs.
For security leaders, the survey’s central message is that adoption, workforce planning and governance are linked. Organizations may be increasing agent use while keeping human approval in the loop, and they are reporting changes to analyst responsibilities rather than a uniform move to remove L1 roles. The data offers a snapshot of surveyed organizations, not a forecast of how quickly autonomy or staffing models will change. Teams evaluating agent deployment can use the reported priorities as areas to examine: performance in their own environment, traceability, enforceable scope limits and a way to reverse actions.