The AI Shortcut Cybersecurity Leaders Should Avoid - eSecurity Planet
Security leaders must avoid using artificial intelligence to provide answers before analysts develop their own judgment, a new report warns.
Artificial intelligence is transforming how cybersecurity teams operate, offering speed and automation that can handle repetitive tasks. However, the technology also presents a significant risk: it may short-circuit the development of critical judgment in future defenders. Leaders must ensure AI is used to support, not replace, the hands-on learning process that builds expertise.
The core issue is that cybersecurity is adversarial work. Analysts need to understand how attackers think and learn to interpret small anomalies. This requires practice, not just access to answers.
AI reshaping workforce structures
The impact of AI on the workforce is already evident. Research from SANS indicates that 74% of cybersecurity teams are changing their structures and role assignments due to the technology. Entry-level roles in Security Operations Centers (SOCs) and for security analysts are among the most affected.
The pressure on professionals is intensifying. A study by ISSA found that 68% of cybersecurity professionals believe their jobs have become harder over the past two years, with nearly half considering leaving their current roles.
The danger of premature answers
Allowing AI to supply answers before analysts form their own conclusions is a dangerous shortcut. Analysts learn by working through messy evidence, making a call, and finding out whether that call was right. If AI gives them the answer too early, they may start to review conclusions before they know how to form their own.
A recent incident illustrates this risk. An experienced analyst used AI automation to investigate a large amount of logs. At the end, the analyst had a feeling something was still amiss. It turned out the automation lacked access to a whole class of logs. A less experienced analyst might have concluded the investigation too soon with a less reliable conclusion.
Protecting the learning layer
Many security careers are built through hands-on repetition. Junior analysts investigate phishing emails, triage alerts, and review suspicious activity. This work, while not always exciting, builds the pattern recognition and judgment needed for complex decisions. It also teaches people how to communicate and present a recommendation to business stakeholders.
Organizations must protect this learning layer. AI can assist with much of the work, but leaders need to be clear about where human review still matters. A better model is to have the analyst review a message, write the first recommendation, explain why the alert matters, and then compare their reasoning with the AI output.
AI works best as a coach when it checks the analyst’s reasoning after the analyst has formed an initial view.
Designing workflows for capability
Security leaders need to treat AI as a workforce design issue. Each AI workflow alters what people practice and how they build confidence in their decisions. Leaders must identify what people still need to learn and where human judgment matters most.
Cybersecurity already faces a significant skills challenge. Many organizations receive plenty of applications but struggle to find candidates with the right experience for emerging roles, particularly in AI security and governance. Leaders cannot hire their way out of this problem and need to build talent inside their organizations.
AI should become part of that development model. It can help junior employees improve faster by explaining incidents and surfacing context. However, this outcome will not happen automatically simply because a company bought another AI-enabled tool.
Beyond efficiency to capability
Companies that succeed with AI will use it to eliminate low-value work while protecting the experiences that build future talent. That starts with identifying what people still need to learn and how junior employees become capable decision-makers.
Security leaders need to ask what AI can take off the team’s plate and what their newest analysts still need to experience. Those analysts still need to make the call, test their reasoning, and learn from the result. AI should sharpen that instinct, not short-circuit it.