AI agents just made cybersecurity 25x faster — and the humans are still in charge.
The AI Agents Automation Breakthrough in Cyber Defence
Sophos, one of the world's leading managed detection and response (MDR) providers, has deployed OpenAI's Daybreak model to slash threat investigation time by a staggering 96%. What once took analysts hours now takes minutes — and that gap is the difference between a contained incident and a full-blown breach.
Even more striking: 52% of MDR cases are now handled autonomously by AI agents, with human analysts stepping in only when genuine judgement calls are needed. That's not AI assisting humans — that's AI doing the heavy lifting while humans supervise the outcomes.
What Daybreak Actually Does (and Why It's Different)
Daybreak is OpenAI's model purpose-built for high-stakes, reasoning-intensive workflows — the kind where a wrong answer has real consequences. In Sophos's pipeline, it ingests raw threat telemetry, correlates signals across endpoints, and produces structured investigation summaries that analysts can act on immediately.
The key design principle here is human oversight preserved. Sophos didn't remove analysts from the loop — it removed the tedious triage work that was burning them out and slowing response times. Analysts now focus on adversarial decision-making, not log-sifting. That's a model worth understanding for any organisation deploying AI agents in sensitive domains.
If you want to understand how multi-agent systems like this are architected, the Inside the Swarm course breaks down how AI agent coordination actually works under the hood.
What This Means for Learners
This story is a masterclass in responsible AI deployment — and it's directly relevant to anyone building AI literacy right now. The Sophos case shows that the most powerful AI implementations aren't about replacing humans wholesale; they're about identifying which tasks AI can own and which require human judgement.
Understanding AI agents — how they're scoped, how they escalate, and how oversight is built in — is rapidly becoming a core professional skill. Whether you work in security, operations, legal, or finance, the same pattern applies: AI handles volume, humans handle nuance. Our AI Agents course is the right place to start building that mental model.
The 96% figure will age quickly — the real skill is knowing how to design systems that achieve it safely in your own context.
