Research ·
AIOps: from buzzword to operating model
In 2017 Gartner described a new kind of IT operations platform built on big data and machine learning. Nearly a decade later, AI sits in the monitoring console, the service desk and, increasingly, in agents that act on their own. Here’s how it got here, what’s working, and what CTOs are planning next.
On this page: Example dashboard · Maturity check · Sources
Stops are evenly spaced, not drawn to scale. Dashed stops are predictions.
2017
The name
Gartner’s April 2017 research described “algorithmic IT operations”: platforms that collect operational data, store it, analyze it and visualize it, connected to service management and automation tools. It grew out of the IT operations analytics tools of the early 2010s. Today the industry reads AIOps as “AI for IT operations”, and the core idea is the same: let software find the signal in the noise so people can act faster. [1]
Share of global enterprises Gartner predicted would have strategically implemented an AIOps platform by 2019 (2017 prediction). [1]
Source: Gartner, Apr 11, 2017 [1]
Gartner’s early advice still holds: AI in operations has to be adopted gradually, and starts with monitoring before any automated action. See the 2021 guidance in the next chapter. [2]
2021
Two kinds of AIOps
By 2021 the market had split. Domain-centric tools applied AI inside one area, such as network or application monitoring. Domain-agnostic platforms pulled events from everywhere and correlated them. Gartner’s guidance that year: roadmaps run three to five years, payback takes several years, and the safest first automated actions are low-risk ones that people have approved in advance. [2]
Source: RTInsights, Apr 21, 2021 [2]
The noise funnel
Illustrative
Correlation turns 1,860 alerts into 14 incidents: about 1 for every 130 alerts.
Illustrative example of what event correlation does, for one day in a mid-size organization. Real results depend on data quality and topology. Bar widths show the stages, not the exact ratios. See the full event management pipeline →
Alert list
2023-2024
Generative AI arrives
Generative AI changed what AI could do for operations staff. It could write: summarize an incident, draft a knowledge article, explain a log, answer a question in plain language. Gartner named democratized generative AI a top infrastructure and operations trend for 2024 and pointed to content generation, knowledge discovery and conversational AI as the useful starting points. [3] [4]
Organizations piloting or in production with generative AI (Gartner poll of 1,400+ executives, Sep 2023). [3]
Source: Gartner, Dec 5, 2023 [3]; Gartner, Date not stated [4]
The service desk becomes the proving ground
The service desk is where AI has paid off first. In Gartner’s late-2025 survey of 782 infrastructure and operations leaders, IT service management was the most common place they reported AI wins. Analysts now treat AI as core to service management: classifying and routing tickets, suggesting resolutions, generating knowledge and answering users through virtual agents. Gartner even published a dedicated Hype Cycle for AI in IT service management in 2025. [5]
I&O leaders whose AI wins occur in IT service management (Gartner survey of 782, late 2025; published Apr 7, 2026). [5]
Source: Gartner, Apr 7, 2026 [5]
Remaining respondents not shown.
Source: ITSM.tools, Sep 16, 2025 [6]
Source: ITSM.tools, Sep 16, 2025 [6]
Forrester’s caution: AI needs high-quality knowledge sources to avoid made-up answers, and shouldn’t be bolted onto old workflows. [7]
Source: Forrester, Date not stated [7]
One ticket, two ways
Illustrative
The same user ticket, “Can’t connect to the VPN since this morning”, handled without and with AI. Minutes are invented to explain the idea.
Without AI 96 min
With AI 19.5 min
Both bars use the same scale: the full width is 96 minutes.
See every step
| Step | Without AI | With AI |
|---|---|---|
| Classify | 6 min | 0.5 min (auto) |
| Route | 25 min | 1 min (auto) |
| Find similar issues | 20 min | 1 min (suggested) |
| Suggest fix | 30 min | 10 min (agent applies suggested fix) |
| Resolve | 15 min | 5 min |
| Write knowledge article | Often skipped | 2 min (drafted, agent approves) |
| Total | 96 min | 19.5 min |
Illustrative example.
2024-2026
The reality check
Expectations ran ahead of results. In 2024 Gartner placed intelligent automation for infrastructure and operations in the Trough of Disillusionment. In 2025 Gartner reframed the core category as event intelligence and urged leaders to separate AIOps hype from achievable value. In 2026 its survey found that only 28% of infrastructure and operations AI use cases fully met their return-on-investment expectations, with the biggest disappointments in auto-remediation and self-healing infrastructure. [8] [9] [5]
Source: Gartner, Sep 18, 2024 [8]; Gartner, Mar 10, 2025 [9]; Gartner, Apr 7, 2026 [5]
Source: Gartner, Apr 7, 2026 [5]
I&O AI use cases that fully met ROI expectations (Gartner, Apr 7, 2026). [5]
I&O heads naming cost optimization as their top goal for AI; lack of budget (50%) and integration difficulty (48%) were top challenges (Gartner survey of 253, published Oct 29, 2025). [10]
Network professionals who fully trust their network data quality for AI; 59% use AI features from their network management vendors but only 35% call AI-driven network management completely successful (EMA survey of 458, Jan 20, 2026). [11]
2025-2026
Agents
The newest step is AI that acts: agents that open the ticket, gather the evidence, run the approved fix and report back. Gartner named agentic AI a top infrastructure and operations trend for 2026, and in December 2025 placed it at the peak of inflated expectations. Most organizations are still exploring. [12] [13]
Source: Gartner, Dec 11, 2025 [12]; Gartner, Dec 10, 2025 [13]
Approximate values as reported by Gartner.
Source: Gartner, Dec 10, 2025 [13]
Over 40% of agentic AI projects will be canceled by end of 2027 (Gartner prediction, Jun 25, 2025), due to costs, unclear value or weak risk controls. [14]
15% of IT application leaders were considering, piloting or deploying fully autonomous agents; 13% strongly agreed they had the right governance (Gartner survey of 360, Sep 30, 2025). [15]
Source: Gartner, Jun 25, 2025 [14]; Gartner, Sep 30, 2025 [15]
An AIOps-assisted operations view
Illustrative dashboard. Example data, not a real system.
Incidents, last 7 days
Day 1: 31 incidents. Day 7: 14.
Correlated incidents
- Checkout latencyProbable cause: database connection pool exhaustedLinked change: pricing release, 7:30 p.m.
- VPN login failuresProbable cause: expired certificate on a gatewayLinked change: none found
- Storage array warningsProbable cause: capacity threshold on one volumeLinked change: backup schedule edit
Service desk queue
| Ticket | AI-suggested category | AI-suggested fix | Action |
|---|---|---|---|
| Can’t connect to the VPN | Network access | Reset the user’s VPN profile | |
| Laptop won’t install an update | Endpoint | Clear the update cache and retry | |
| Shared mailbox is full | Raise the mailbox limit one tier | ||
| Report page times out | Application | Restart the reporting service | |
| New starter needs access | Access request | Apply the standard role template |
The Approve buttons are visual only in this example. In practice a person approves each action.
2028-2029
What analysts predict
Predictions, not measurements. Dashed bars mark predictions. The first and third bars come from the same source; the second from a different one.
Sources: Gartner, Nov 17, 2025 [16] (60% by 2028, and only 15% of agents fully autonomous within seven years); Gartner, Jan 21, 2026 [17] (70% by 2029).
For CTOs this is a planning item for years, not a single purchase. The pattern is consistent: start where AI has proven itself (the service desk, event correlation, summarization), keep people approving actions, and invest in the data that everything else depends on.
The data underneath
What makes AIOps work
Every source points the same way: AI is only as good as the data it learns from. Gartner predicts organizations will abandon 60% of AI projects that aren’t supported by AI-ready data through 2026, and found 63% of organizations lacked, or weren’t sure they had, the right data management practices for AI. For operations, that data is your telemetry, your event stream and your CMDB: what exists, who owns it, and what depends on what. Gartner also warns that short-lived cloud assets make traditional discovery and dependency mapping obsolete; keeping the CMDB accurate now needs continuous discovery and service mapping. Even researchers point to the shortage of reliably labeled incident data as the main obstacle for anomaly detection. [18] [19] [20]
Source: Gartner, Feb 26, 2025 [18]; Gartner, Feb 21, 2025 [19]; arXiv, Not stated [20]
The AIOps stack
- People approving actions
- Service desk and automation
- Correlation and analytics
- Topology and CMDBCMDB Health check →
- EventsFrom signal to ticket →
- Telemetrymetrics, logs, tracesOpenTelemetry →
Weak data at the bottom limits everything above it.
Where are you on the AIOps ladder?
Five rungs, from collecting data to acting with guardrails. Six questions, about a minute. You get a scorecard with steps to fix each gap. Your answers stay in your browser unless you choose to send a project request.
- 5 · Act with guardrailsApproved automated actions, agents with human oversight.
- 4 · Predict and assistAnomalies flagged, fixes suggested, service desk AI.
- 3 · ContextualizeTopology/CMDB and traces attached.
- 2 · CorrelateEvents grouped, noise reduced.
- 1 · CollectMonitoring in silos.
The maturity check needs JavaScript.
Planning for this in your organization?
King Tech collaborates on custom solution architecture: we work alongside your team to design an approach that fits your systems, your constraints and your timeline. Contact King Tech to start the conversation.
Sources
- Gartner, “Press release: algorithmic IT operations drives digital business”, Apr 11, 2017.
- RTInsights, “Reporting on Gartner’s 2021 Market Guide for AIOps platforms”, Apr 21, 2021.
- Gartner, “Press release: top trends impacting infrastructure and operations for 2024”, Dec 5, 2023.
- Gartner, “Article: highlights from the Gartner IT Infrastructure, Operations & Cloud Strategies Conference”, Date not stated.
- Gartner, “Press release: AI projects in infrastructure and operations stall ahead of meaningful ROI”, Apr 7, 2026.
- ITSM.tools, “The State of AI in ITSM (Mid-2025)”, Sep 16, 2025.
- Forrester, “The Future Of ESM Is Intelligent”, Date not stated.
- Gartner, “Press release (infrastructure and operations automation, Hype Cycle placement)”, Sep 18, 2024.
- Gartner, “Research document (event intelligence)”, Mar 10, 2025.
- Gartner, “Press release: survey of infrastructure and operations leaders adopting AI to cut costs”, Oct 29, 2025.
- Enterprise Management Associates (EMA), “Research: AI adoption gap in network operations”, Jan 20, 2026.
- Gartner, “Press release: top trends impacting infrastructure and operations for 2026”, Dec 11, 2025.
- Gartner, “Press release: IT Infrastructure, Operations & Cloud Strategies Conference 2025, Las Vegas, day 2 highlights”, Dec 10, 2025.
- Gartner, “Press release: over 40% of agentic AI projects will be canceled by end of 2027”, Jun 25, 2025.
- Gartner, “Press release: survey of IT application leaders on fully autonomous AI agents”, Sep 30, 2025.
- Gartner, “Press release: IT Infrastructure, Operations & Cloud Strategies Conference 2025, London, day 1 highlights”, Nov 17, 2025.
- Gartner, “Article: AI for infrastructure and operations”, Jan 21, 2026.
- Gartner, “Press release: lack of AI-ready data puts AI projects at risk”, Feb 26, 2025.
- Gartner, “Research document (CMDB, discovery and dependency mapping)”, Feb 21, 2025.
- arXiv, “Preprint 2304.04661 (anomaly detection and labeled incident data)”, Not stated.
Figures are as published by each source; predictions are not measurements. Illustrative examples are labeled. Trademarks belong to their owners.
Planning something like this? Talk to us. · Observability services