Powered by Splunk
Agent AI makes IT and security teams dramatically more efficient. But it also eliminates the apprenticeship that has long trained experienced operators.
As organizations automate more of the work once performed by junior analysts and engineers, they face a workforce design challenge as much as an architectural design one: how to build the next generation of professionals when artificial intelligence takes over the work they were once taught to do.
What does a small workforce do
Over two decades, the path to becoming a world-class SecOps analyst, SRE or NetOps engineer has been repeated.
Testing for false positives. Hunt through dashboards for context. Reading magazines at 2 in the morning worked well. The industry viewed this work as drudgery, and in many ways it was.
But it also served as an apprenticeship.
An analyst’s thousands of hours poring over traffic rules have built intuition that makes them invaluable when a real attack comes. This intuition was not taught in a single course or captured in a runbook. It is collected through exposure, pattern recognition, failure and escalation. Over time, people gain deep analytical experience.
However, agent AI is now beginning to automate tasks that once served as training grounds for this practice. This is not a reason to slow down. The challenge came at a cost. The exhaustion was real. Organizations should use agents to reduce labor wherever they can.
At the same time, as we eliminate this apprenticeship cycle, we need to provide operators with something better in its place. How organizations approach this issue today will determine the future winners.
Organizations that are intentional about this will produce skilled operators to succeed in the next decade. Organizations that do this may end up with faster systems today, but with fewer people who understand them deeply enough to operate them tomorrow.
When automation removes accountability
There is a second dimension to this conversation that gets less attention than it should.
The difficulty of learning in regulated environments is part of the level of accountability. From SOX to PCI DSS to HIPAA to NIS2, frameworks assume a chain of human judgment behind a control decision.
Auditors do not interview models. They interview people who can explain why the system did what it did, why the decision was correct, and whether the proper controls were in place.
The risk may not be immediately apparent when the pool of experts who can explain this chain begins to thin. Control can still be passed. The workflow can still be executed. The dashboard may still appear green.
But the core organizational memory is starting to run out.
It’s not just a tool problem. It is also a workforce skill and design challenge. For organizations moving quickly to adopt an agent, the risk is closer than many think.
Building human experience to drive AI
When we lose some of the accountability layer to agents, people will move into a different management role. Managing an agent system means implementing automated safeguards that adapt to and ensure non-deterministic agent behavior.etc agents behave appropriately in circumstances that no one fully expects. This means designing escalation criteria that capture the right anomalies without overwhelming people with the wrong ones. This means implementing dynamic tools, alerts, and processes to review machine decisions to detect drift, bias, and reasoning failures that no individual case would detect.
The ability to appreciate and respond to these exceptions requires judgment based on years of experience, learning to recognize the pattern produced by the old apprenticeship model.
So the manpower question and the architecture question are now the same question. If we expect humans to operate increasingly autonomous systems, we need intentional ways to help humans manage the scale and speed of AI systems while building the intuition and judgment in the human operators required to do the job.
In the age of artificial intelligence, the most valuable platforms will simply not automate most tasks. They will help people become more capable, more reliable and more important as the systems around them become faster and smarter.
This means that organizations must invest in a complete ecosystem of expertise for operators: communities that disseminate shared experiences, certifications or other evidence that makes expertise visible, and human-centered explanations and validations in AI, as well as learning paths that develop skills. Reinforcement is an architectural design choice
Human empowerment is an important part of the conversation around the practical use of artificial intelligence. However, without a deliberate strategy to back it up, it risks becoming a meaningless statement, as it could mean anything.
Empowerment for agent systems cannot simply be a conceptual requirement. It should be a set of design choices made about how systems behave. An agent system that empowers human operators and develops their professional skills does four things:
1. It exposes the reasoning behind the data generation
Every recommendation an agent makes must be traceable to the data it considers, the logic it applies, and the origin of the data it uses. Operators who can see thoughts develop judgment about when to trust it. Operators don’t just give results.
2. Rank authority by trust and influence
Familiar, low-risk patterns can be handled autonomously. New situations or moves with a significant blast radius should be raised by default. The boundary should be clear and configurable by the teams who own the results.
3. Treats controversy as a correction signal
When a seasoned engineer trumps an agent, they do more than just disagree. They fix the model with a judgment that is not in the model: a fragile dependency, a quirk in the environment, a constraint that the data never sees. A system that registers a cancellation but ignores the reasoning behind it learns nothing from when a person knows better.
4. Captures resolutions as cross-domain information
How an incident is resolved is rarely a lesson left in a lane. A SecOps incident can expose an ITOps vulnerability. A network problem can be caused by a business impact. If this connection is only inside a closed ticket, the next team to hit it starts from scratch. Resolutions should propagate across domains, not die where they are issued.
These are not desirable qualities. They are testable product opportunities. Leaders evaluating agent systems must be able to determine where these capabilities reside, what happens when they fail, and how operator skills improve after implementation.
The next advantage is that human and artificial intelligence can scale together
An important design point for AI systems to be practical, reliable, and scalable is for AI to work deeply with and empower human operators.
So the age of agency is not a story of replacing people. It’s a story about redesigning the systems people work in so that these operations can happen at machine speed and scale, while also enhancing the human experience. Not at the expense of each other, but together.
This result is not given. This will happen when leaders treat operator development as a priority, not an afterthought. To achieve this, agent systems must be designed to expose thinking, capture learning, and feed it back to people in ways that build skills and careers, rather than eroding both.
Agents will continue to get smarter and faster. The ability of the operators who work with them to learn and evolve will determine whether the next decade of digital resilience is something organizations truly own or something they rent from a dwindling pool of expertise.
Learn more about how Cisco Data Fabric powered by the Splunk Platform helps teams accelerate agent operations.
Kamal Hathi is SVP and CEO of Splunk, a Cisco company.
Sponsored articles are content produced by a company that paid for the post or has a business relationship with VentureBeat and is always clearly marked. Contact for more information sales@venturebeat.com.
