> ## Content Index
> Fetch the complete content index at: https://www.cybrsecmedia.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# Is Security Automating Away Its Own Apprenticeship?
- URL: https://www.cybrsecmedia.com/security-is-automating-away-its-own-apprenticeship/
- Published: 2026-11-18T13:34:28.000Z
- Updated: 2026-11-18T13:54:11.000Z
- Description: AI can now do the grunt work that turned junior analysts into experts. Unless security leaders deliberately cultivate that training ground, the profession will run short of people who can tell when the machines are wrong.
- Author: George V. Hulme
- Tags: Security Automation, CYBR.SEC.Community, blog

Dann Huff has produced Keith Urban's records for two decades. So when the Nashville producer told an audience last week that some AI-assisted song demos sounded, at first listen, "better than anything I've ever done," he wasn't selling anything. He was pointing to a challenge facing the music and cybersecurity professions.

Huff was onstage with Urban and Anna Griffin, Commvault's chief market officer, at SHIFT, the data-protection company's customer conference. He described the "farm league" of the music business: Songwriters have long hired young session players to cut demos, the rough recordings used to pitch a song. The pay was modest, and the stakes were low. But the work put players in situations they'd never meet practicing alone, and next to people who could later hire them. In Huff's experience, musicians moved up because their playing improved. They also moved up because the people they had worked with moved up and brought them along. Now a songwriter can make a convincing demo without hiring anyone. Huff likes the speed. He also knows who no longer gets the gig.

Cybersecurity is making the same trade, but with high stakes. Security teams are handing AI agents the alert triage, documentation, and reporting that have long been the job of entry-level analysts. That makes each team faster today. It also removes the work through which beginners learn the judgment the profession will need to check those agents tomorrow. Every security chief who automates junior work is making a sensible choice for their own organization. Taken together, those choices are draining the talent pool everyone hires from.

Call it the expertise paradox: The tools that let novices perform like experts make them less likely to become experts.

[Subscribe to the CYBR.SEC.Media newsletter](#/portal/signup/free)

## **The grunt work was the curriculum**

ISC2, the largest professional association for security practitioners, asked hiring managers last year what they assign to new hires. Documentation topped the list, followed by alert and event management and report writing. That's exactly the work AI agents now do well.

Consider a single alert: someone logged into an employee's account from an unfamiliar country at 3 a.m. A junior analyst working that case under supervision must decide whether the evidence points to a stolen password or a sales rep on a hotel network. The analyst must notice what's missing, such as a travel request or a recent password reset. The analyst must also weigh the cost of being wrong in either direction. Lock the account, and you might stop an intruder, or you might cut off a sales director in the middle of closing a deal.

An agent can close that ticket in seconds. The organization gets its answer, and the analyst learns nothing. The ticket wasn't just valuable for its answer. It was also valuable for the reasoning it forced the analyst to do.

## **The profession is telling on itself**

In May, ISC2 surveyed 856 security professionals who use AI at work. Fifty-six percent said AI had reduced the need for entry-level positions. Fifty-three percent said it was creating new kinds of entry-level work. Read quickly, that looks like disagreement. It isn't. The old jobs are disappearing faster than anyone is defining the new ones.

The rest of the survey makes the case for worry. Eighty-nine percent of respondents said they had seen AI recommendations lead to wrong outcomes at their organizations. Half said that when this happens, a human is held accountable. And 62% said AI hasn't reduced the need for foundational security skills.

The machines make mistakes, people take the blame, and fundamentals still matter. Yet the profession is shrinking the jobs where people learn those fundamentals. Respondents split almost evenly on whether hands-on learning opportunities have already shrunk, 37% to 36%. That near-tie should worry security leaders. It means the damage hasn't shown up clearly yet, and it may not until it's too late to fix cheaply.

## **Jevons won't save the bench**

The hopeful counterargument dates back to 1865\. William Stanley Jevons observed that more efficient steam engines didn't reduce Britain's coal consumption. They made coal cheap enough for new uses, and total consumption rose. Applied to security, the argument goes like this: cheaper investigation means more of it, and more machine output means more work for humans to review. Radiology is the usual example. AI was supposed to replace radiologists, and instead their [numbers have grown](https://ajronline.org/doi/full/10.2214/AJR.24.31357?ref=cybrsecmedia.com), though radiologists themselves dispute how [much AI had to do with it](https://www.benwhite.com/radiology/radiology-isnt-an-example-of-jevons-paradox-yet/?ref=cybrsecmedia.com).

 The ISC2 data partly supports the optimists. About two-thirds of respondents said they now spend more time deciding whether to trust AI recommendations (65%) and validating what AI produces (63%). AI oversight is becoming a job.

The optimists miss who gets that job. Spotting a confident, well-formatted wrong answer takes exactly the judgment that junior work used to build. Left to default, managers will give the review work to senior analysts, because they're the only ones who can do it reliably. Jevons explains why the work will exist. He doesn't say who will be trained to do it. That part is a management decision, and right now most organizations are making it by default.

## **Design the new farm league with intention**

Turning oversight into apprenticeship doesn't take a new product. It takes rules that a security manager could put in place this quarter.

Give juniors real cases before they see the AI's verdict. Require them to tie each conclusion to evidence, list the other explanations, and detail what would change their minds. Then compare their reasoning with the machine's. Where live data is too sensitive, use incident replays, anonymized recreations of past attacks. Give juniors seats on containment decisions, the calls about when to cut a laptop off the network or end a user's session, and on the reviews after an incident. Don't leave them with whatever the agent couldn't handle. 

Measure the program the way you measure any security control: if junior development isn't incentivized, it will be the first thing cut in a busy quarter.

Urban described a small version of this. He said he doesn't use AI to write music. He does use Moises, an app that separates a finished recording into its individual instruments, to study how other artists build arrangements. The software does the separating. Urban does the learning. Security teams should hold their AI to the same standard and choose tools that show their reasoning, so a junior analyst can examine it, argue with it, and learn from it.

## **The market will fix this too late**

The other comforting argument is that scarcity is self-correcting. If senior analysts become rare, salaries rise, and employers start training again. Maybe. But it takes years to develop a senior analyst. By the time pay signals a shortage, the people who should have been trained will have gone into other careers.

 This is Garrett Hardin's tragedy of the commons. Each herder gains from adding one more animal to a shared pasture, and everyone shares the cost of overgrazing. Security's pasture is its supply of experienced practitioners. Any one company can rationally hire only seasoned analysts and let someone else train them. When every company does that, no one trains anyone.

Fixing a commons takes people who are willing to give up some private advantage. CISOs should keep a fixed share of junior seats even when automation makes those seats look optional. Boards should ask about talent pipelines as seriously as they ask about uptime. Professional bodies like ISC2 should track the size of the training ground, not just sentiment about it.

Urban's rule for his own work was "Don't outsource any skill that you're not willing to lose yourself." Security is now outsourcing the skill it can least afford to lose: producing its own experts. Someone on a 3 a.m. incident call in 2035 will have to overrule a machine that sounds certain whether that person knows how is being decided in this year's budgets.

[![](https://storage.ghost.io/c/ab/67/ab676516-71e3-473d-8f73-9e0692f5aaee/content/images/2026/11/2027-Events-Banner-45.png)](https://www.cybrsecmedia.com/conference/)