“Most teams start the same way: one agent, one long prompt telling it to act like a Level 1 analyst, and a pile of tools bolted on.” That assessment from Tim Leehealy, co‑founder and president of Strike48, captures the friction at the heart of a rapid change in security operations: agents are now widespread, but teams are still learning how to use them safely and effectively.
Strike48 survey of 534 security and IT leaders
Strike48 polled 534 security and IT leaders at organizations with 200 or more employees that maintain a security operations function in the UK, Canada and the US to compile The State of Agentic Security H2 2026. The headline: 87% of respondents are redesigning their security operations centers (SOCs) because of AI agents. Adoption has jumped quickly — 74% said they have deployed agents in their SOC, up from 36% in May — and almost half (49%) report using agents across multiple use cases.
L1/Tier‑1 roles are being reshaped, not eliminated
The survey directly addresses a common fear: that AI will remove the entry rung from cybersecurity careers. Rather than wholesale cuts, organizations report redeployment and reskilling. Half (49%) of respondents are hiring for new skills such as AI detection engineering and automation. Two‑fifths (42%) are retraining L1 analysts to supervise agents, 38% are shifting L1s into investigation roles, and 30% are expanding teams with additional entry‑level analysts.
Strike48’s analysis notes these patterns will “allay common concerns in the industry that AI will take the roles of entry‑level SecOps employees, eliminating a vital career on ramp.” The data in the report show workforce change is occurring through new job definitions and training rather than through simple replacement.

The cyber insurance questionnaire just landed. Now what?
SOC 2, HIPAA, insurance renewals - someone has to own security strategy. Nubivance provides fractional CISO leadership without the full-time salary.
Get a security leadTrust limits agent autonomy in SOCs
Widespread deployment has not translated into unrestricted operational autonomy. Only 30% of teams say they allow agents to act with high or full autonomy. Nearly half — 47% — require humans to execute actions or preapprove them, and 23% let agents act only within defined boundaries.
Strike48’s respondents identified five specific requirements that increase trust in agent autonomy (respondents could select multiple): proven accuracy in their environment over time (57%); a full audit trail of every action and the data behind it (53%); technical enforcement that the agent cannot exceed limits, not just instructions telling it not to (46%); a precisely defined scope for each agent (44%); and a reliable rollback or undo for any action (40%). These governance and technical controls appear to be the levers teams are using to move from cautious adoption toward greater autonomy.
Swimlane findings on analyst skill development
Complementary evidence from a September Swimlane report underscores a tension beneath the adoption numbers: a portion of analysts feel their career development is affected. The Swimlane report found that 24% of SOC analysts feel agent technology has limited their ability to build skills, and one in 10 expect fewer opportunities for junior analysts to gain experience. Those perceptions sit alongside the Strike48 finding that many organizations are actively retraining and hiring for new roles — indicating divergent experiences across teams and deployments.
What this means for technologists, enterprise leaders, and L1 analysts
- Technologists and security teams: Expect to design agents as narrowly scoped tools embedded in deterministic workflows rather than single, open prompts. Tim Leehealy argues teams that reach autonomy “break the work into narrow jobs and fixed, deterministic workflow steps.”
- Enterprise procurement and leaders: Governance controls will be decisive procurement criteria. The survey’s top trust requirements — audit trails, technical enforcement of limits, rollback capabilities and proven accuracy — are explicit specifications buyers will likely request from vendors.
- Entry‑level analysts and trainers: The data show mixed outcomes but a clear pathway: retraining for supervision, investigation roles, and new AI‑adjacent skills are already in motion, with nearly half of respondents hiring for those competencies.
The record in Strike48’s H2 2026 report is straightforward: agents have moved from experiment to mainstream, but teams are cautious about handing them the keys. Adoption has accelerated, job descriptions are changing, and trust — built through accuracy, auditable actions, enforced limits and rollback mechanisms — remains the gating factor for autonomy. As organizations redesign SOCs, the decisive question is not whether agents will be used, but whether governance and narrow, deterministic workflows will be put in place at scale to make them both useful and safe.
https://www.infosecurity-magazine.com/news/secops-changing-l1-role-as-agent/




