AI agents are cybersecurity firms' newest employees

As cyber threats continue to evolve and grow more sophisticated, security firms are grappling with an escalating workload and a shortage of skilled analysts. To address these challenges, many organizations are turning to a new workforce solution: AI agents. Unlike traditional generative AI tools that depend on user inputs, AI agents are designed with specific roles and trained to manage complex, multi-step workflows. A recent McKinsey survey from 2025 revealed that 62% of companies are currently experimenting with AI agents. The trend is particularly pronounced in cybersecurity, where a study by ISC2 found that 30% of professionals have integrated AI security tools into their operations. Many of these tools are evolving to function like agents capable of handling tasks previously performed by human analysts. Companies are now tasked with implementing these AI-driven innovations across various teams. Early implementations have shown promising results, but the technology does have its limitations, leading to questions about how quickly AI agents can adapt to high-stakes environments and what this means for future job roles. Huntress, a cybersecurity platform, has deployed nearly 20 AI agents within its Security Operations Center (SOC), which oversees threat alerts for 240,000 clients. According to Eric Stride, the Chief Security Officer, these agents automate investigations that were once the responsibility of a 50-member SOC team. In the identity threat detection process, an AI agent identifies suspicious activities such as unusual login attempts and unauthorized inbox rules. This triggers an orchestration agent, which acts as an AI supervisor to delegate tasks and activate sub-agents that gather data, analyze actions, and detect evasion tactics. This system can differentiate between malicious and benign activities, escalating complex cases to human analysts when necessary. The efficiency of this system is striking. Stride indicated that processes that previously took 20 to 30 minutes can now be completed in mere minutes, reducing the workload for analysts by a remarkable 90% for over a third of investigations and generating approximately 10,000 incident reports monthly. Analysts are now able to focus on more intricate attacks, with Stride likening their enhanced capabilities to having an 'Iron Man suit' in their arsenal. Similarly, DNSFilter has integrated an AI agent into its customer support team, which consists of less than ten engineers. This AI agent manages all Tier 1 tickets, categorizing incoming requests by complexity. It resolves straightforward issues using internal documentation while escalating more complex problems to human staff. This has drastically improved response times, allowing the AI agent to handle 60 requests weekly compared to the 35 tickets typically managed by a human. Despite these advancements, AI agents still face notable challenges. Stride noted that while agents excel at repetitive tasks, they struggle with ambiguous requests and may yield inconsistent results. They are less adept at detecting intricate threats such as ransomware attacks and require human oversight for high-stakes decisions. Pruitt from DNSFilter also pointed out limitations, including the agent's reliance on internal documentation, which once led to a significant error in advising a customer. Economically, the benefits of AI agents are compelling. Pruitt stated that an AI agent costs around $15,000 to $16,000 annually, equating to the output of two full-time support engineers, resulting in potential savings of $200,000 per year for the company. This financial incentive is prompting DNSFilter to reconsider its hiring strategies, particularly for entry-level positions, as the capabilities of AI agents improve. Both Huntress and DNSFilter view AI agents as a means to enhance productivity without increasing headcount. "We're striving to make our team of 150 perform like a team of 500," Pruitt noted, with aspirations of achieving this by the end of the year.