In depth
The "Response" half of the acronym matters as much as detection. A modern EDR can isolate an endpoint from the network at the kernel level (allowing the EDR backend to still reach in but blocking everything else), terminate processes, quarantine files, kill network connections, run custom remediation scripts, capture memory dumps and disk artefacts, and roll back ransomware-encrypted files (where supported). These response actions are typically callable from the SOAR playbook, which is how a phishing-click on a payroll laptop becomes an automatically-isolated host within seconds of the detonation.
EDR's strengths are visibility and response speed — the agent sees what a SIEM that only consumes Windows Event Logs cannot — and the ability to do retrospective threat hunting across weeks of historical telemetry when a new IOC drops. The weaknesses are agent footprint (CPU and memory overhead, plus the not-zero risk of agent-induced kernel panics, as the 2024 CrowdStrike incident illustrated), licence cost, and coverage gaps on unmanaged devices, BYOD, OT and certain server workloads where agent installation is not feasible.
EDR is now a near-universal control. PCI DSS v4.0 Requirement 5.2.2 (anti-malware), SOC 2 CC7.1, ISO 27001 Annex A.8.7 (Protection against malware), HIPAA Security Rule 164.308(a)(5)(ii)(B), and most cyber-insurance underwriting questionnaires effectively require it. See adversary simulation services for validating EDR coverage under realistic conditions.