AI agents may help security operations centres investigate threats, but Crogl chief executive Monzy Merza says they could initially bring more work and confusion. His practical point: teams should define the outcomes they need before shopping through vendors’ feature lists.
Watch Desk analysis
What happened
In a 16 November interview with Digital Journal, Merza argued that adopting AI in security operations is as much an organisational challenge as a technical one. He urged teams to decide what work AI should do and what should remain with people, and to consider operational questions such as where data goes and how work continues if a model provider has an outage.
Merza also warned that security teams need to understand how their own AI tools and agent harnesses handle suspicious links, permissions and files. A system used to investigate an attack could help achieve the intended outcome, leak data or, in the worst case, carry out an attacker’s instructions. The Digital Journal interview also discusses sovereignty, model failover and the skills future security teams may need.
Why it matters
Security teams are being asked to adopt tools that can make decisions probabilistically while still protecting sensitive systems and information. Merza’s argument is that buying a longer feature list will not settle the harder questions: what a team wants to achieve, who is accountable, and what happens when the model or service is unavailable.
Those questions are especially practical for smaller teams and managed security providers. Merza predicts that routine technical work may be increasingly automated, while customers will continue to value help managing risk. That is his outlook, not proof that a particular job or service is already being replaced.
Our read
The useful advice here is to start with the work, not the sales demo. Before handing an agent a role in a security operation, teams need to decide what it may access, what a person must approve and how the work carries on when the model does not.
An agent can add another set of hands. It can also add another set of permissions to manage, which is less thrilling but rather more important.
What to watch
- Whether security teams define measurable outcomes before choosing AI tools.
- How organisations handle data sovereignty and service outages in AI-assisted operations.
- Which actions teams reserve for human approval, and what skills agent operators need.
Discussion spark: Should security teams automate more investigation work now, or wait until they can clearly define what an agent may access and do?
Sources and evidence
- Q&A: AI is coming for the SOC, are security teams ready? – Digital Journal (16 November 2026, 21:21 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.