Agentic artificial intelligence (AI) significantly increases the efficiency of IT and security teams, while automating many routine tasks that historically trained junior operators. Presented by Splunk, a Cisco Company, the piece argues this is as much a workforce-design problem as an architecture problem: organizations must figure out how to cultivate the next generation of experts when AI takes over the tasks that once taught them.
Why junior work mattered
For decades, the career path to becoming a top SecOps analyst, SRE, or NetOps engineer ran through repetition: triaging false positives, hunting dashboards for context, and reading logs at 2 a.m. that often proved benign. Although much of that work was tedious and contributed to burnout, it also served as an apprenticeship.
The thousands of hours an analyst spent observing traffic patterns built the intuition that proved invaluable during real incidents. That intuition was not learned in a single course or captured in a runbook; it was accumulated through exposure, pattern recognition, failure, and escalation.
Agentic AI is now beginning to automate those same training-ground tasks. That alone is not a reason to slow adoption: reducing toil cut costs and lowered burnout. But removing the apprenticeship loop forces organizations to provide something better in its place. How they handle that trade-off now will shape who succeeds in the coming decade.
When automation hollows out accountability
In regulated environments—frameworks like SOX, PCI DSS, HIPAA, and NIS2—the drudgery of apprenticeship is part of the accountability layer. Auditors interview people, not models; they need humans who can explain why a system acted as it did and whether appropriate controls were in place.
If the population of professionals able to explain that chain thins, the immediate operational signal may be muted—the control might still pass and dashboards may remain green—but organizational memory begins to erode. This is therefore not only a tooling issue but also a workforce-skill and design problem, and for organizations rapidly adopting agentic systems, the risk is nearer than many assume.
Building human expertise to govern AI
Losing part of the accountability layer to agents means humans must assume a different governance role. Governing agentic systems requires automated guardrails that adapt to non-deterministic agent behavior and ensure agents behave properly under unanticipated conditions. It requires escalation criteria that surface real anomalies without overwhelming humans with false positives. It demands dynamic tools, alerts, and review processes to detect drift, bias, and reasoning failures that single cases would not reveal.
Evaluating and responding to these exceptions requires judgment built over years of experience—pattern recognition that the old apprenticeship model produced. Hence the workforce question and the architecture question are now the same: expecting humans to govern autonomous systems means creating deliberate pathways that let people manage AI’s scale and speed while building the intuition and judgment needed to do that work.
Practical elements to grow operator expertise
Agentic systems that truly empower human operators and grow professional skillsets do four concrete things:
- Expose reasoning, with data lineage
- Every agent recommendation should be traceable to the data it considered, the logic it applied, and the provenance of inputs. Operators who can see reasoning develop judgment about when to trust it; those given only conclusions do not.
- Tier authority by confidence and impact
- Familiar, low-risk patterns can be handled autonomously; novel situations or actions with meaningful blast radius should escalate by default. Boundaries must be explicit and configurable by the teams that own the consequences.
- Treat disagreements as correction signals
- When an experienced engineer overrides an agent, they supply judgment the model lacked: a fragile dependency, an environmental quirk, or a constraint absent from the data. A system that records overrides but ignores the reasoning learns nothing from the moments humans know better.
- Capture resolutions as cross-domain knowledge
- Incident resolutions rarely stay in one lane. A SecOps incident may expose ITOps weaknesses; a network issue may link back to business impact. If resolutions remain trapped in closed tickets, the next team starts from zero. Resolutions should travel across domains.
These are not merely aspirational goals but testable product capabilities. Leaders evaluating agentic systems should identify where these capabilities exist, what happens when they fail, and whether operator skill improves after deployment.
The next advantage is when human and AI scale together
For AI to be practical, trusted, and effective at scale, it must operate deeply alongside—and empower—human operators. The agentic era is therefore not about replacing people but redesigning the systems people operate so work can occur at machine speed while human expertise grows in parallel.
That outcome is not guaranteed. It will occur only where leaders treat operator development as a priority rather than an afterthought. Agentic systems must be designed to expose reasoning, capture learning, and route work back to humans in ways that build skills and careers rather than erode them.
Agents will continue to get faster and smarter. Whether organizations own the next decade of digital resilience or rent it from a shrinking pool of expertise depends on whether the operators working alongside those agents can learn and grow in lockstep.
Learn more
Cisco Data Fabric powered by the Splunk Platform is presented as a solution to accelerate agentic operations. Kamal Hathi is Senior Vice President and General Manager of Splunk, a Cisco Company.
This article is sponsored content.
