The AI landscape doesn't move in one direction — it lurches. Some techniques leap from experiment to table stakes in a single quarter; others stall against regulatory walls, technical ceilings, or organisational inertia that no amount of hype can dislodge. Knowing which is which is the hard part. The State of Play cuts through the noise with a rigorously maintained index of AI techniques across every major business domain — classified by maturity, evidenced by real-world adoption, and updated daily so you always know where you stand relative to the field. Stop guessing. Start knowing.
A daily newsletter distilling the past two weeks of movement in a domain or two — delivered to your inbox while the index updates in the background.
Each dot marks the weighted maturity of practices within a domain — hover for a brief summary, click for more detail
AI that executes predefined incident response playbooks automatically, containing threats and preserving evidence. Includes SOAR platform automation and orchestrated containment; distinct from automated remediation in IT ops which restores service rather than containing threats.
Incident response automation is at an inflection point between proven SOAR maturity and emerging agentic systems. Traditional platforms (XSOAR, Splunk SOAR, MS Sentinel) established the category with 40–90% manual effort reductions but achieved only 40–55% alert automation in practice versus claimed 95%, limited by playbook maintenance and governance complexity. Since Q1 2026, major vendors (AWS CIRT, Azure SRE Agent, CrowdStrike Falcon Fusion, Arctic Wolf Aurora) shipped autonomous IR systems demonstrating 3–10 minute investigations (vs. 20–40 human minutes), 70–95% alert triage automation, and 85–90% MTTR reduction at scale. Agentic IR deployment surged 412% (8% to 41% adoption in 2025). The practice is operationally mature and mainstream, but two critical challenges define the tier ceiling: (1) governance calibration—agentic systems require immutable audit logs, per-decision records for forensic defensibility, human approval gates for high-impact actions, and incident response runbooks specifically for agent failure modes (silent degradation, evidence fragmentation, model-version losses, evidence collection gaps in compromised environments); (2) adoption paradox—despite high deployment rates (65% of orgs adopted AI agents), only 10% report excellent value (SOC-CMM 2026 survey, ~200 SOCs), with structural gaps from isolated tool silos (triage agent unaware what detection engineers silenced; threat hunting agents ignoring threat intel findings). Real-world incidents document weaponization via prompt injection (Unit 42: insider using agentic IR to stage unauthorized data exfiltration) and 53% of organizations experiencing AI agent scope violations (CSA 2026). Practitioner frameworks now specify containment escalation (tool revocation → queue drain → shadow mode) and incident-specific playbooks (prompt injection, agent tool-call escalation, model extraction). The practice is proven viable and mainstream, but success remains contingent on end-to-end IR orchestration, forensic auditability, immutable decision artifacts, and organizational acceptance that agentic systems require permanent governance infrastructure—not point-tool deployment.
Vendor ecosystem and deployment maturity accelerated in June 2026 with multi-agent platforms and cloud-native GA products. Product releases: CrowdStrike Falcon Fusion (agentic SOAR, Charlotte AI) achieved 70% customer efficiency improvement, automated 900 false positives (75 hours/month saved) at a healthcare org; AWS CIRT (fully managed automated IR with AI investigative agent) now GA with bidirectional ITSM integrations and multi-source correlation; Microsoft published AI incident response playbook (June 2026) with structured telemetry collection (Purview Unified Audit Log → Sentinel) and specific thresholds (50+ Copilot events/hour = outlier; one-hour correlation patterns for credential-theft + deployment-write sequences); Azure SRE Agent (GA March 2026) delivers 40–70% MTTR reduction with dual autonomous/review modes; Arctic Wolf Aurora orchestrates 300+ agents automating 90% of Tier-1/2 tasks. Named deployments confirm operational viability: NTT Data reduced end-to-end investigation from 154 to 12 minutes (94.9% accuracy); Druva multi-agent Bedrock deployment achieved 68% IR automation (30–60 day investigations to minutes); fintech MCP-based autonomous IR achieved <5 min MTTR (from 45 min baseline). Market size: SOAR $1.87B (2025, 18.6% CAGR), incident response market forecast $243.7B by 2035. Agentic IR deployment surged 412% (8% to 41% adoption in 2025). NIST SP 800-61 Revision 3 (April 2025) explicitly endorses alert triage and playbook automation.
Critical governance deficits and attack surface risks constrain adoption despite deployment maturity. Real-world security gaps: Unit 42 documented insider using agentic IR for unauthorized data exfiltration via prompt injection; recent attack analysis shows 72-minute attacker timeline (initial access to data exfiltration), requiring automated behavioral-sequence playbooks for containment before manual analysis concludes. MCP attack surface: 13,000+ public servers, 43% vulnerable to command injection, 30% to SSRF, 22% to path traversal—enforcing governance prerequisite on autonomous tool-call execution. Organizational readiness: CSA survey (600+ orgs): 53% experienced AI agent scope violations, 47% had incidents, 97% expect major AI agent incident within 12 months. SOC-CMM 2026 survey (~200 SOCs): only 10% excellent AI value despite deployment surges (+55–145% YoY), attributing gaps to isolated silos (triage agent unaware of what detection silenced; threat hunting agent ignoring threat intel). Governance framework emerging: immutable per-decision audit records (prompt, context, tool call, approval state), containment escalation levers (tool revocation → queue drain → shadow mode), forensic preservation (model version, prompt hash, idempotency keys), incident-response-specific playbooks for agent failure modes (silent degradation, evidence fragmentation, evidence collection gaps). D3 Morpheus and multi-agent platforms (Stellar Cyber, Torq HyperSOC) demonstrate 95% alert triage (<2 min), but adoption barriers persist: playbook maintenance burden remains (legacy SOAR achieves 40–55% automation vs. claimed 95%), integration drift costs (18-min autonomous repair vs. 4–6 week manual rebuild). The practice is operationally proven and mainstream, but success requires permanent playbook governance, scope discipline, immutable forensic telemetry, and organizational acceptance that agentic systems demand incident-response-specific failure-mode runbooks and governance infrastructure rather than point-tool adoption.
— Market transition from SOAR to agentic AI with Ponemon research: AI/automation orgs save $1.9M per breach and shorten lifecycle 80 days. NIST SP 800-61 Rev3 (April 2025) explicitly endorses playbook automation.
— SANS June 2026 survey on AI adoption in detection/response workflows, workforce evolution, and tool adoption trends—independent analyst benchmark on IR automation maturity.
— Operational playbook for agent IR containment with escalation levers (tool revocation, queue drain, shadow mode); requires compensating transactions, idempotency keys, and model-version pinning for forensic accuracy.
— Unit 42 analysis of 4× year-over-year attack compression (initial access to data exfiltration in 72 min). 87% of incidents required cross-platform correlation; recommends automated playbooks for documented behavioral sequences.
— Autonomous SOC migration metrics: 95% alerts triaged + L2-investigated in <2 min, 18-min integration drift MTTR (vs 4-6 week industry baseline); eliminates playbook maintenance burden.
— Microsoft's June 2026 AI incident response playbook (Purview Unified Audit Log → Sentinel) organizes AI investigations with specific thresholds (50+ Copilot events/hour = anomaly; one-hour correlation patterns for credential theft + deployment writes).
— AWS CIRT GA: fully managed automated incident response with AI investigative agent correlating CloudTrail/IAM/cost data producing actionable timelines within minutes; bidirectional ITSM integrations.
— Multi-agent autonomous IR using MCP (triage, investigation, remediation, documentation agents) with named fintech outcome: MTTR reduced from 45 min to <5 min; durable workflow orchestration with human checkpoints.