Agentic CTI is the use of AI agents to carry out parts of the cyber threat intelligence cycle on their own: collecting relevant reporting, extracting indicators and behaviors, matching them to the organization's priorities and turning them into hunts, detections or briefings. Analysts direct the work and make the final calls on assessments.
It applies the same idea as the agentic SOC to the intelligence side of security operations.
CTI needs agents because the volume of threat reporting far exceeds what a small team can read, and most of the reading is sorting rather than analysis.
A CTI analyst might face dozens of vendor reports, government advisories, sharing-community posts and feed updates in a single day. Most are irrelevant to the organization. Of the relevant ones, extracting the indicators, mapping behaviors to MITRE ATT&CK and checking whether anything has already been seen internally is careful but routine work. Agents can do that first pass, so analysts spend their time on judgment: what does this mean for us, and what should we do?
PIRs keep agents focused by defining exactly what questions matter, so agents can filter thousands of reports down to the handful that answer one.
A requirement like "Which ransomware groups are targeting healthcare organizations in North America, and what initial access methods are they using this quarter?" gives an agent a clear test for every incoming report. Without PIRs, an agent collects everything and the noise problem just moves. With them, each new report is tagged to a requirement, matched against internal data and either surfaced or set aside with a reason.
Agentic CTI meets the SOC at the point where intelligence becomes action. When an agent extracts new indicators from a report, they flow through automated integration and enrichment at ingest. When it maps a group's behaviors, it can draft hunt queries for the SOC to run against historical data. And when SOC agents see activity tied to a known actor, that feeds straight back into the intelligence picture. This is the loop described in unifying CTI and SOC teams, running at machine speed.
Can AI agents write threat intelligence assessments?
They can draft them, but analysts should own the final assessment and its confidence level. Agents are strongest at collection, extraction and matching.
How do agents avoid hallucinating intelligence?
By working from curated, sourced intelligence and citing the source of every claim. Unsourced statements should be treated as unverified. See our page on agent hallucination.
Do small CTI teams benefit from agents?
Often the most, because agents take on the processing work that would otherwise consume a one- or two-person team.
Anomali applies agentic AI to curated intelligence in ThreatStream Next-Gen, so agents work from scored, sourced intelligence and can match it against live and historical telemetry.
Request a demo to see agentic CTI in Anomali.