Presented by Splunk
Agentic AI is significantly boosting the efficiency of IT and security teams, yet it is also disrupting the traditional apprenticeship model that has long cultivated skilled operators.
As companies increasingly automate tasks formerly handled by junior analysts and engineers, they face a dual challenge involving workforce and architecture design: how to nurture future experts when AI takes over the roles that once trained them.
The Role of Junior Workforce
For the past twenty years, the journey to becoming a top-tier SecOps analyst, SRE, or NetOps engineer has relied on repetitive tasks.
These include sorting through false positives, navigating dashboards for context, and examining logs in the early hours of the morning that often proved harmless. While this work was often seen as tedious, it also functioned as a crucial training period.
The countless hours spent analyzing traffic patterns honed the intuition necessary for responding effectively to genuine attacks. Such intuition isn’t something taught in a single course or outlined in a manual; it’s developed through exposure, pattern recognition, and dealing with failures and escalations. This is how individuals accumulate deep analytical expertise over time.
However, agentic AI is beginning to automate these foundational tasks. This shift should not be seen as a detriment. The tedious work was costly and led to burnout. Organizations should employ agents to alleviate such burdens wherever possible.
As we phase out this traditional training cycle, it’s essential to offer operators something more valuable in its place. The strategies organizations adopt today will define their future success.
Those that approach this thoughtfully will cultivate skilled operators for the coming decade, while those that neglect this may end up with advanced systems today but fewer individuals who deeply understand and can manage them tomorrow.
The Impact of Automation on Accountability
Another aspect of this issue often overlooked is its effect on accountability.
In regulated industries, the monotonous work of training is part of the accountability framework. Regulations like SOX, PCI DSS, HIPAA, and NIS2 presume a chain of human judgments behind control decisions.
Auditors do not engage with models; they speak with people who can explain a system’s actions, justify decisions, and confirm the presence of appropriate controls.
As the number of professionals capable of explaining these processes dwindles, the risks may not be immediately apparent. Controls might still appear compliant, workflows operational, and dashboards normal.
Yet, the organization’s memory begins to erode.
This is not merely a tooling issue but also a challenge in workforce skills and design. For organizations rapidly adopting agentic AI, this risk is nearer than many might anticipate.
Developing Human Expertise for AI Governance
As agents take over parts of the accountability process, humans will assume new governance roles. Managing an agentic system involves setting automated safeguards that can adapt to unpredictable agent behavior and ensure agents act correctly, even in unforeseen scenarios. It also requires establishing escalation criteria that highlight genuine anomalies without burdening humans with irrelevant ones. Additionally, it involves setting up dynamic tools, alerts, and processes to monitor machine decisions for drift, bias, and reasoning errors not evident in single cases.
Responding to these exceptions demands judgment, built up through years of experience and pattern recognition once gained through traditional training models.
Therefore, the questions of workforce and architecture design merge into a single inquiry. To govern increasingly autonomous systems, we must create intentional pathways that enable individuals to manage AI systems’ scale and speed while developing the intuition and judgment needed for such work.
In the AI era, platforms offering the most value will not only automate tasks but also enhance human capabilities, credibility, and importance as surrounding systems become faster and smarter.
This necessitates investment in a comprehensive ecosystem of expertise for operators, including communities to share best practices, certifications or proofs to make expertise visible, human-oriented explanations and verifications in AI, and learning paths that build skills. Empowerment becomes an architectural design decision.
Human empowerment is a crucial topic in the practical application of AI. Without a deliberate strategy to support it, it risks becoming an empty phrase.
Empowerment in agentic systems cannot remain a mere concept; it must be a set of design choices embedded into system behaviors. An agentic system that empowers human operators and fosters their professional development should do four things:
1. Reveal Reasoning and Data Origins
Every recommendation an agent makes should be traceable to the data considered, the logic applied, and the inputs used. Operators who can see this reasoning develop judgment on when to trust it. Operators who receive only conclusions do not.
2. Differentiate Authority by Confidence and Impact
Low-risk, familiar patterns can be managed autonomously. Unusual situations or actions with significant consequences should automatically escalate. The boundaries should be explicit and configurable by the responsible teams.
3. Use Disagreements as Correction Signals
When an experienced engineer overrides an agent, they provide more than just a disagreement. They correct the system with insights the model lacks: a delicate dependency, an environmental peculiarity, or a constraint not visible in the data. A system that acknowledges the override but ignores the reasoning behind it learns nothing from the moment a human had superior knowledge.
4. Share Resolutions Across Domains
The resolution of an incident often offers insights beyond its immediate context. A SecOps incident might reveal an ITOps vulnerability, or a network issue might have business ramifications. If these connections remain confined to closed tickets, future teams start from scratch. Resolutions should circulate across domains instead of being confined to where they were logged.
These are not mere aspirations; they are tangible product capabilities. Leaders evaluating agentic systems should pinpoint where these capabilities reside, what happens when they fail, and whether they enhance operator skills post-deployment.
The Cooperative Growth of Humans and AI
For AI systems to be functional, reliable, and scalable, they must work closely and empower human operators.
The agentic era isn’t about replacing humans; it’s about redesigning systems to enable operations at machine speed and scale while simultaneously enhancing human expertise. This collaboration should be mutually beneficial rather than competitive.
This outcome is not guaranteed. It will be realized only where leaders prioritize operator development rather than treating it as an afterthought. Achieving this requires agentic systems to be purposefully designed to expose reasoning, capture learning, and reassign tasks to humans in ways that enhance skills and careers rather than diminish them.
Agents will continue to become increasingly intelligent and quicker. The ability of operators who work alongside them to learn and progress in sync will determine if the next era of digital resilience is something organizations genuinely possess or merely lease from a dwindling pool of expertise.
Discover more about how Cisco Data Fabric powered by the Splunk Platform is aiding teams in advancing agentic operations.
Kamal Hathi is SVP and GM of Splunk, a Cisco Company.
Sponsored articles are content produced by a company that either funds the post or has a business relationship with VentureBeat, and they’re always clearly marked. For more information, contact sales@venturebeat.com.
