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-enhanced monitoring of application and network performance to detect degradation, predict issues, and recommend optimisation. Includes APM anomaly detection and network traffic analysis; distinct from AIOps alerting which correlates across systems rather than monitoring specific layers.
AI-enhanced application and network performance monitoring is standard operating infrastructure for IT organizations at scale. Adoption has crossed the mainstream threshold—Gartner projects 50% of enterprises with distributed data architectures will deploy observability tools by 2026, up from 20% in 2024. New Relic's 6.6M-user dataset shows AI-enabled teams deploying 80% faster and resolving incidents 25% faster than non-AI peers, confirming both business impact and production maturity. The market is consolidating: Datadog maintains dominance with $3.43B FY2025 revenue and 32,700 customers generating trillions of telemetry data points daily. The core practice is no longer debated—it is the execution landscape that remains contested. Organizations deploy three or more monitoring solutions despite consolidation intent, face pervasive alert fatigue (engineers receive 50 alerts/week with only 2-5% actionable), and confront a critical emerging gap: traditional APM tooling remains infrastructure-centric and structurally fails for AI workloads. LLM and agentic AI applications demand observability for semantic correctness, hallucination detection, and token-cost tracking—capabilities that current APM platforms do not provide. The practice has reached operational maturity; the frontier is specialization for AI systems.
Tier-1 vendors—Datadog, Dynatrace, and New Relic—control market momentum through scale and product velocity. Datadog's 32,700-customer base generates trillions of telemetry data points hourly, with 84% of customers using two or more products and cross-product customers generating 15x higher revenue; the platform's GA MCP Server integrates observability data into AI coding agents. New Relic's 6.6M-user Intelligent Observability Platform demonstrates AI adoption at scale: teams using AI capabilities ship 5x more frequently (453 vs 87 deployments per day), achieve 2X alert correlation, and resolve incidents 25% faster than non-AI counterparts. Dynatrace's Davis AI offering auto-adaptive thresholds, forecasting, and native GenAI observability integrations (Bedrock, Azure AI Foundry, OpenAI) with confirmed case studies reducing issue identification time by 90% and MTTR by 95% in healthcare deployments.
Practitioner adoption sentiment reveals a maturity paradox. Grafana Labs surveyed 1,300+ practitioners in March 2026: 91-92% see value in AI-assisted anomaly detection, RCA, and dashboard generation—diagnostic capabilities are table-stakes—but only 49% rate autonomous remediation as valuable, reflecting organizational caution on unguarded automation. Organizations continue managing 3+ monitoring tools despite 97% expressing intent to consolidate, driven by tool-specific champion ownership and migration cost. The APM market reached $10.7B in 2025 and grows at 12.6% CAGR to $12.06B (2026), projected $20.19B by 2030—growth fueled almost entirely by AI/ML integration and the shift from reactive threshold alerting to proactive anomaly detection. However, critical adoption barriers remain permanent: alert fatigue (engineers receive ~50 alerts weekly with only 2-5% requiring action), vendor lock-in (94% of IT professionals concerned), and a fundamental specialization gap. Traditional APM platforms cannot monitor AI systems: OpenTelemetry captures infrastructure surface but not semantic correctness, hallucination rates, model drift, or token-cost attribution—deficiencies that drive organizations to build custom observability layers for LLM and agentic workloads rather than relying on vendor solutions.
— Identifies critical NPM gap: traditional SNMP polling misses microbursts in AI-driven networks; requires real-time streaming telemetry and packet-level analytics for AI traffic patterns—documents monitoring blind spot for AI workloads.
— Q1 2026 earnings analysis: NRR re-accelerated to 120%, 56% of customers using 4+ products (up from 51%), hyperscalers deploying Datadog internally for AI research—structural evidence of platform consolidation and AI workload adoption.
— Independent research: non-AI customer revenue growth mid-20s% YoY (refuting concentration thesis), hyperscaler training wins represent TAM expansion, but management's guidance conservatism flags concentration risk—balanced critical assessment of durability.
— Survey of 300 engineering professionals: 89% use AI tools but only 8% apply AI to observability; 72% predict AI mission-critical in 2-3 years but require verifiable outputs and human-in-the-loop workflows—quantifies adoption barriers and organizational caution.
— Institutional analysis of Q1 2026: all-time record ARR sequential adds, new logo average land size doubled YoY, RPO +51% YoY; 35% use 6+ products (up from 28%)—validates platform consolidation and expansion acceleration.
— Q1 2026: $1.01B revenue (+32% YoY), 4,550 customers at $100K+ ARR, 6,500+ AI-integrated customers generating ~80% of ARR; GPU Monitoring and Bits AI Security Agent launched—quantifies production APM deployment across enterprise base.
— GigaOm analyst consensus: AI/ML-powered anomaly detection and LLM assistants are primary market differentiators; unified AI-driven platforms consolidating visibility, analytics, and automation across hybrid/multicloud environments as dominant architecture.
— Omdia study of 80 global CSPs: 79% seeing AI services traffic, 47% using AI in production networks with 48% faster troubleshooting/RCA and 40% lower OpEx; 56% expect autonomous networks within 3 years—production deployment with documented outcomes.
2017: Market emergence ($1.6B, 20.7% CAGR) with vendor AI feature launches (Dynatrace Davis, New Relic Applied Intelligence); research and critical assessment highlighted both promise and maturity limitations.
2018: Major vendor expansions with Datadog Watchdog (ML anomaly detection), Dynatrace Davis platform integration, and emerging vendors (Bwtech NetWarden); industry research confirmed strong market demand, with 90% of enterprises unable to meet SLAs due to inadequate monitoring visibility. APM/NPM was moving from early adoption to mainstream vendor feature parity.
2019: Cloud platform integration accelerated with AWS CloudWatch Anomaly Detection and Open Distro Elasticsearch plugins, Dynatrace Davis extended to Azure and Kubernetes. Real-world deployments emerged (Kapital Bank, cloud migration case studies). Adoption forecasts positive (75% plan automation increase, 58% expect AI/ML impact) but actual adoption remained ~13%, constrained by complexity and tool fragmentation.
2020: APM/NPM consolidated into production-grade mainstream offerings. Datadog unified platform matured with Watchdog expanding from APM to infrastructure monitoring; Dynatrace enhanced Davis for Kubernetes-scale deployments; vendors demonstrated real-world ROI (Carta fintech case, ERT DevOps automation, Rack Room Shoes e-commerce uplift). Specialized network vendors (NVIDIA UFM Cyber-AI, Broadcom DX NetOps) entered market with AI-driven NPM solutions. Survey data showed strong adoption signals (89% of CIOs accelerating digital transformation, 70% identifying automation as priority), but visibility gaps persisted (enterprises averaged 10 tools but achieved full observability on only 11% of environments). Practice moved from early mainstream to mature production deployment stage, though organizational barriers (tool sprawl, skill gaps) continued to limit enterprise-wide adoption.
2021: APM vendor products and cloud migrations demonstrated measurable business value. Datadog cases (Seven.One Entertainment 100% adoption with 78% cost savings, Neto cloud migration success) confirmed real-world enterprise deployment patterns. Dynatrace extended Davis AI error detection capabilities. Industry surveys showed growing acknowledgment of monitoring tool inadequacy (90% of practitioners report traditional tools insufficient for modern cloud stacks). Critical analysis emerged noting that traditional APM struggles with cloud-native complexity, signaling evolution toward broader AIOps. Practice at mature production stage with strong enterprise deployment evidence, though complexity barriers remained.
2022-H1: Enterprise deployments accelerated with strong ROI evidence. ArcXP achieved 15% MTTR improvement through Datadog APM; BT consolidated 16 legacy platforms into Dynatrace, targeting £28m savings. Datadog extended Watchdog with Log Anomaly Detection and Root Cause Analysis. Industry research (ESG/Splunk) quantified value: observability leaders reduced downtime costs by 90% and launched 60% more innovations. Dynatrace maintained market leadership (12 consecutive Gartner APM quadrant wins). Tool sprawl and skill gaps remained primary adoption barriers despite demonstrated ROI.
2022-H2: Vendor case studies validated APM/NPM production deployments across diverse sectors (energy, finance, industrial IoT), while research and analyst reports documented persistent adoption barriers. IETF research highlighted fundamental challenges in AI/network management integration: NP-hard optimization, data quality, and explainability gaps limiting real-world implementation. EMA survey found 84.8% of orgs unable to detect network issues pre-impact and 53% false-positive rates in alerts, indicating tool maturity gaps despite vendor claims. New Relic's 1,614-respondent survey confirmed the complexity-adoption paradox: 78% view observability as business-critical, but only 27% achieved full-stack observability and 82% maintain 4+ fragmented monitoring tools. Observability leaders' demonstrated ROI and innovation velocity continued to widen the gap from organizations struggling with implementation complexity and tool sprawl.
2023-H1: Tier-1 vendors (Datadog, Dynatrace) extended AI capabilities into new layers. Datadog announced Log Anomaly Detection and Root Cause Analysis (April), plus Watchdog for Real User Monitoring (May), consolidating AI across observability stack. Dynatrace maintained Gartner leadership (ranked #1 across all six APM use cases), while Perform conference data showed 30% application and 211% auxiliary workload growth—signaling accelerating Kubernetes adoption complexity. Mid-market vendors (ManageEngine OpManager) added ML adaptive thresholds and RCA, expanding AI-augmented monitoring beyond premium market segment. Industry maturity deepened: ESG/Splunk 2023 survey positioned observability beyond early adoption, with market leaders demonstrating reduced outage rates and stronger business outcomes. However, adoption remained constrained: enterprises continued managing 4+ fragmented tools, only 27% achieved full-stack coverage, and network observability gaps persisted (84.8% unable to detect issues pre-impact, 53% alert false-positive rates). Practice remained at mature deployment stage with growing vendor capabilities but persistent implementation barriers.
2023-H2: Vendor ecosystem continued AI expansion with focus on emerging use cases. Datadog and New Relic launched dedicated AI stack monitoring features (Datadog AI integrations, New Relic AI Monitoring), signaling market recognition of observability requirements for LLM and vector database applications. Dynatrace advanced Davis AI with hypermodal consolidation (predictive, causal, and generative AI), demonstrating evolution toward more sophisticated autonomous problem diagnosis. Academic research accelerated on network performance topics with peer-reviewed advances in traffic forecasting (LSTM with transfer learning) and anomaly detection (CNN-GAN hybrid models), validating ML technical foundations for NPM. Market maturity surveys continued to show structural adoption barriers despite expanding vendor capabilities: enterprises remained constrained by tool fragmentation, skill gaps, and the expanding complexity of cloud-native and AI application monitoring.
2024-Q1: Vendor focus shifted to operationalizing AI for emerging workloads. New Relic released AI Monitoring (GA) with auto-instrumentation for LLM frameworks and an AI Assistant for natural language observability querying (March 2024). Dynatrace extended Davis AI to custom data streams (StatsD, Telegraf, Prometheus) and launched OpenPipeline for petabyte-scale telemetry processing with 5-10x performance improvement and 30% deduplication (February 2024). Market research (Elastic survey, 500+ respondents) confirmed 94% ROI sentiment from observability investments, with generative AI and tool consolidation as top priorities. Despite vendor innovation, critical APM deployment barriers persisted: data silos, skill gaps, tool sprawl, cost justification, and scalability overhead continued to constrain enterprise-wide full-stack adoption. Practice remained at mature production stage with expanding capabilities but persistent organizational implementation friction.
2024-Q2: Generative AI and hybrid cloud complexity emerged as central vendor focus. New Relic and Dynatrace announced AI-powered automation and extended LLM/vector database monitoring (June 2024). Market surveys confirmed broad AI adoption momentum (75% of enterprises) but persistent implementation barriers: only one-third of NPM practitioners satisfied with tools despite 70% managing cloud, only 10% achieved full observability (Logz.io), and MTTR continued increasing (82% over 1 hour). Practitioner assessments identified specific limitations in anomaly detection (false positives, baseline drift) limiting confidence in fully automated monitoring. Practice remained at mature production stage with widening gap between vendor technical capability and enterprise operational scaling.
2024-Q3: Tier-1 vendor leadership confirmed with quantified ROI evidence from independent analyst studies. Forrester TEI analysis demonstrated 267% ROI and $5.1M NPV from New Relic deployments with 40% IT time savings and 70% MTTR reduction. Dynatrace extended Davis AI across hybrid multicloud (Nutanix integration). Generative AI application monitoring began production deployment (OneStream customer case). However, critical cost and sustainability concerns emerged: Gartner warned 30% of AI projects abandoned by 2025 due to escalating costs and ROI uncertainty; vendor lock-in risks and persistent tool fragmentation (53% alert false-positive rates) continued limiting enterprise-scale adoption despite mature technical capabilities.
2024-Q4: APM/NPM vendor ecosystem consolidated around AI-driven features with formal industry standards advancement. AWS established AI/ML monitoring as formal best practice (TELCOPERF04-BP01) in Well-Architected Framework with specific guidance for SageMaker, Kinesis, and CloudWatch. Enterprise adoption remained steady but fragmented: 60%+ of enterprises deployed at least one APM solution; major market grew to USD 9.94B with 11.5% CAGR (2025-2034); 70% of large US enterprises used APM tools; 52% of new 2024 solutions incorporated AI-driven anomaly detection, yet tool sprawl persisted (45% of organizations managing 5+ monitoring tools, only 25% achieving full-stack observability). Adoption barriers remained structural: integration complexity (evidenced by production deployment issues requiring runtime-level fixes), vendor lock-in concerns, and implementation overhead continued constraining enterprise-scale adoption despite proven ROI from leading implementations.
2025-Q1: Vendor innovation focused on GenAI observability and predictive operations. Dynatrace extended AI observability for generative AI applications with LLM model analytics, guardrails, and multi-model tracing (January); Datadog expanded Watchdog to detect faulty Kubernetes deployments (January). Independent case study validation emerged: Dynatrace Perform 2025 conference data showed healthcare provider achieving 90% reduction in problem identification time and 95% MTTR reduction. Media/entertainment industry survey showed 60% adoption of AI monitoring with 296% ROI. Research elevated critical perspectives: IETF draft articulated fundamental challenges in coupling AI with network management (acceptability, explainability, security gaps), while practitioner analysis identified traditional APM limitations for LLM applications (quality metrics, cost/token tracking, model behavior monitoring), indicating emerging specialization requirements beyond APM consolidation.
2025-Q2: Autonomous remediation and AI stack specialization emerged as vendor differentiation strategy. Datadog announced Bits AI agents (SRE, Dev, Security) with 50% incident resolution reduction and 1000+ monthly PRs (June); Microsoft Azure expanded AI-powered investigations and health models at Build 2025 (May); Dynatrace continued extending AI observability for GenAI applications. Adoption metrics showed accelerating enterprise deployment: New Relic's 85,000-customer dataset revealed 30% quarterly growth in AI monitoring usage with 92% increase in unique LLMs, validating production maturity. Analyst recognition confirmed tier-1 leadership: Forrester Wave AIOps Q2 2025 positioned Datadog and Dynatrace as Leaders. However, vendor lock-in crystallized as explicit adoption barrier—proprietary instrumentation and context propagation created migration complexity despite OpenTelemetry standardization; 60-98% cost reduction potential in alternatives highlighted pricing and auto-scaling unpredictability. Tool sprawl, false positives (53% in network monitoring), and baseline drift persisted as constraints despite mature technical capabilities.
2025-Q3: Market adoption acceleration continued with quantified growth and major cloud platform commitment. New Relic's September 2025 survey of 1,700 IT professionals confirmed AI monitoring adoption grew to 54% (up from 42% in 2024), with full-stack observability cutting outage costs in half. Datadog's Q2 earnings showed 28% YoY revenue growth driven by AI, with 4,500+ customers using AI integrations and AI-native customers representing 11% of revenue (up from 8%). Microsoft Azure released GA tooling for monitoring GenAI applications, demonstrating tier-1 cloud vendor investment in specialized AI application performance monitoring. However, critical deployment barriers emerged: MIT NANDA Initiative research revealed 95% of AI pilot projects fail to deliver measurable financial returns due to integration gaps and learning curves, contextualizing the challenge of scaling AI-driven monitoring adoption. Vendor lock-in and cost unpredictability remained structural constraints, with analyses documenting 40% cost reduction potential available through platform migration—signaling that procurement friction and architectural dependencies continued limiting enterprise-scale deployment despite proven ROI and accelerating adoption in leading-edge organizations.
2025-Q4: Sector-specific adoption metrics validated APM/NPM maturity with quantified ROI across telecommunications and technology, while specialized AI application monitoring emerged as distinct capability gap. New Relic survey data (500+ respondents) showed 74% of telcos and 52% of technology firms deployed AI monitoring, with 10% of telcos reporting 5-10x ROI over measurement period; median outage costs of $2M/hour in telecom and $1M/hour in retail drove observability investment prioritization. Microsoft Azure advanced GenAI monitoring at scale with AI Foundry integration, Dynatrace maintained Gartner market leadership, and new vendors (Kloudfuse) unified traditional and AI observability with natural language query capabilities. Industry adoption trajectory assessment (IBM) documented 86% of tech leaders find traditional methods insufficient, with growth projected from 3% current deployment to 25% by 2026. However, critical capability gaps surfaced: independent analysis identified traditional APM tools (Datadog, New Relic, Dynatrace, Splunk) structurally inadequate for AI workload monitoring, lacking statistical drift detection, model behavior analytics, and cost-per-token tracking required for LLM applications. Vendor lock-in, proprietary instrumentation, and false-positive alert rates (53% in network monitoring) remained endemic constraints despite mature technical capabilities. Practice reached peak operational maturity with proven sector-specific ROI, broad enterprise deployment, and 52%+ AI feature penetration in new solutions, yet permanent structural barriers (vendor dependencies, specialization gaps for AI applications, alert fatigue from false positives) constrained transition to universal enterprise adoption.
2026-Jan: Vendor consolidation advanced toward AI-driven observability with enterprise production deployments demonstrating sector-specific maturity. Dynatrace AI Observability GA (January) integrated native support for Generative AI, LLMs, and agentic workflows with TELUS agentic AI optimization case study; Datadog expanded platform dominance with real-world deployments in critical sectors (Toyota autonomous vehicle/smart manufacturing AI/ML monitoring, TriZetto healthcare claims processing and clinical decision support consolidating $2.3M tool spend). Peer-reviewed research (AnomLocal, PLoS One) validated technical foundations with 87-89% federated and hybrid anomaly detection accuracy, confirming algorithmic maturity. However, analyst research crystallized scaling barriers: Forrester (January 2026) documented only 10-15% of AI projects reaching sustained production, with 60% failing due to integration, data quality, and workflow redesign complexity, contradicting vendor maturity claims. Industry trend reports (IBM, LogicMonitor) identified three drivers reshaping adoption: AI intelligence for monitoring AI systems themselves, observability as cost management tool (55% of leaders lack spending visibility), and OpenTelemetry adoption to mitigate vendor lock-in. Practice at peak technical maturity with sector-specific customer evidence and algorithmic validation, yet sustained structural adoption barriers (integration complexity, cost unpredictability, tool fragmentation, specialization gaps for LLM workloads) maintained the gap between leading-edge deployments and enterprise-wide adoption.
2026-Feb: Vendor platforms extended AI-driven autonomous operations with GA releases cementing ecosystem maturity. New Relic Advance 2026 launched Intelligent Workloads and SRE Agent for autonomous incident management (February); Dynatrace Davis AI advanced anomaly detection (Auto-Adaptive Threshold, forecasting); Microsoft Azure Anomaly Detector documented production limitations (minimum data points, no contextual understanding). Enterprise case studies validated APM/NPM adoption across critical infrastructure: FinTech company achieved 99.99% uptime with 60% MTTR reduction using Datadog Kubernetes monitoring (500+ clusters); multiple deployments (Toyota AGV WiFi, BARBRI cloud migration, retail operations) demonstrated sector-wide maturity. Industry surveys (LogicMonitor, Parallels) confirmed sustained adoption momentum with consolidation headwinds: 96% of VP+ IT leaders expect observability spending to hold/grow; 84% pursuing tool consolidation; 47% prioritize AI for issue detection; yet 94% concerned about vendor lock-in, with only 29% willing to pay more for AI features. Practice at sustained peak maturity with quantified ROI in critical sectors and proven autonomous remediation capabilities, yet consolidation and vendor lock-in concerns crystallizing as primary adoption friction for enterprise-scale deployment.
2026-Apr: Market metrics and practitioner sentiment confirmed mature mainstream adoption with a structural specialization gap for AI workloads.
2026-May: Production case studies deepened APM/NPM ROI evidence: AppFolio achieved 80-90% latency reduction and 300% adoption increase using Datadog LLM Observability on Amazon Bedrock; Modulus Labs consolidated fragmented tooling into Datadog achieving 40%+ MTTR reduction in payment infrastructure; Superset reduced production incidents 80% with Datadog monitoring including token usage and PII detection; AssemblyAI demonstrated deep instrumentation maturity for GPU/multi-cloud AI inference pipelines. Eino's agentic network observability reached GA with 1,500+ production deployments across critical infrastructure (airports, refineries, ports), reporting 90% reduction in troubleshooting time; a Rohde & Schwarz survey of 75 network vendors found 97.4% plan AI/GenAI capabilities, confirming near-universal vendor commitment to AI-driven NPM. Gartner projects 50% of enterprises with distributed data architectures will adopt observability tools by 2026 (vs 20% in 2024); APM market reached $10.7B in 2025, growing to $12.06B in 2026 at 12.6% CAGR. New Relic's 6.6M-user study showed AI-enabled teams deploying 80% more frequently and resolving incidents 25% faster; Datadog Q1 2026 earnings showed FY2025 revenue of $3.43B with 32,700 customers, 42% CAGR (2020-2025), and 10x growth in AI observability data usage, with Bits AI agents now autonomously detecting and remediating incidents. Broadcom DX O2 26.3.1 shipped a Spring GenAI extension providing APM instrumentation for GenAI applications with token economics, latency tracking, and safety filter metrics; AWS CloudWatch Application Signals reached GA as a fully managed APM service. Grafana Labs' survey of 1,300+ practitioners found 91-92% value AI anomaly detection and RCA, but only 49% rate autonomous actions as valuable—diagnostic AI is table-stakes while autonomous decision-making remains contested. Peer-reviewed research confirmed OpenTelemetry's structural inadequacy for LLM observability: it captures infrastructure surface but cannot detect hallucination rates, semantic drift, or token-cost attribution, reinforcing the growing gap between traditional APM and AI-workload monitoring requirements. Mid-May 2026 metrics confirmed platform consolidation momentum: Datadog NRR re-accelerated to 120% with 56% of customers using 4+ products (up from 51%); hyperscalers now deploying Datadog internally for AI research and training divisions. Communications service providers accelerating adoption with 47% using AI in production networks and documenting 48% faster troubleshooting and 40% OpEx reduction—production deployment phase validated. Separately, enterprise surveys documented persistent adoption friction: observability practitioners only 8% apply AI to observability tasks despite 89% using AI in daily workflow; 72% predict AI mission-critical in 2-3 years but require verifiable outputs and human-in-the-loop validation. Network infrastructure strain from AI workloads exposed critical monitoring blindspot: traditional SNMP polling cannot detect microbursts in AI-driven traffic, requiring real-time streaming telemetry and packet-level analytics as emerging monitoring requirement.