AIOps Platforms

Put the 2022 and 2025 vendor lists for this category side by side and the interesting thing is not who scored well. It is who stopped existing.

The Forrester Wave: Artificial Intelligence For IT Operations, Q4 2022 evaluated eleven providers. By the time The Forrester Wave: AIOps Platforms, Q2 2025 came around, Micro Focus had been absorbed into OpenText, OpsRamp had gone to HPE, Splunk had been bought by Cisco, and the field had shrunk to ten. Even the category name had changed.

That is what a consolidating market looks like from the outside. AIOps is not dying, it is being eaten, and understanding which direction it is being eaten from tells you most of what you need to know before shortlisting anything.

What AIOps was supposed to solve

The original problem was alert volume. A large IT estate throws off more telemetry than any human team can read, and most of it is noise. A single failing database can generate hundreds of alerts across dozens of dependent services, and somewhere in that pile is the one alert that explains the other ninety nine.

AIOps applies machine learning to that pile. Correlate related events into one incident. Suppress the noise. Identify the probable root cause. Where possible, trigger the fix without waking anyone.

The measurable promise is mean time to resolution, and the secondary promise is that your on-call engineers stop burning out. Both are real. What has changed is who gets to make that promise.

Why the category is being absorbed

Effective correlation requires access to the underlying telemetry. Not a feed of alerts from someone else's monitoring, the actual metrics, logs, and traces, with the dependency map that says which service calls which.

That requirement quietly settles the competitive question. A vendor that already owns the observability layer starts with the data. A vendor that has to ingest it from third parties starts with a copy, arriving late, missing context. Forrester's Q2 2025 evaluation put weight on exactly this, treating native telemetry access as a differentiator rather than an implementation detail, and flagging vendors that lean too heavily on third-party functionality.

So the observability platforms are winning a category that used to be adjacent to them. AIOps is becoming a capability of the monitoring stack rather than a product you buy separately.

There is a real exception worth naming. Organisations running a genuinely heterogeneous estate, several monitoring tools across several teams with no realistic path to consolidation, still need something that sits above all of it. That is the space where the independents survive, and it is a legitimate architecture rather than a consolation prize.

What The Forrester Wave: AIOps Platforms, Q2 2025 found

Published in April 2025, the evaluation scored ten providers against twenty six criteria across current offering, strategy, and customer feedback, with references drawn from up to three customers per vendor.

Dynatrace took the highest score in current offering, with top marks on seventeen criteria including log management, automated remediation, and pricing transparency, credited to its Davis engine for root cause analysis and continuous dependency mapping. ScienceLogic took the highest score in strategy, on a stated vision of agentic AI leading to autonomous operations, and was positioned for buyers wanting a unified platform driven centrally by automation. Datadog also placed as a Leader, with perfect scores across an unusually wide spread of criteria covering innovation, log management, data governance, infrastructure, APM, service mapping, digital experience monitoring, incident detection, automated remediation, and collaboration.

Other publicly identifiable participants include Splunk, New Relic, Elastic, PagerDuty, BMC, and OpenText. The tenth is not identifiable from public sources, so treat that list as nine of ten rather than the complete field.

Two vendor notes from the evaluation are worth carrying into a buying conversation. PagerDuty, which came to this market from on-call scheduling, was characterised as an operational enrichment layer that uses AI but has no native observability of its own and therefore depends on third-party telemetry. And Splunk's reference customers, while positive on the depth of content and its effect on junior operators, raised the Cisco integration as an open question and noted the ongoing effort required to keep configurations and instrumentation current.

Forrester's own summary of the market described AIOps as mature but still transforming, and made a point of distinguishing horizontal insight across the estate from vertical insight within a single domain. That distinction is the whole argument for buying a platform rather than accepting whatever correlation your monitoring tool ships with.

The agentic turn

Every vendor in this category is now describing a future where the platform does not just identify the problem but resolves it without a human in the loop. ScienceLogic's strategy score rested on precisely that vision.

Treat those claims with the scepticism you would apply to any automation promise in production. Automated remediation is well established for known failure modes with known fixes, which is genuinely valuable and not new. Autonomous handling of novel incidents is a different proposition, and the gap between those two is where most of the marketing sits.

The practical question is not whether a platform can remediate automatically. It is what happens when the automated fix is wrong, how quickly you find out, and how quickly you can roll it back.

Reading the category as a buyer

Start by working out which architecture you are actually in. If your telemetry is consolidated in one observability platform, your AIOps decision is largely made, and the real question is whether that vendor's correlation is good enough or whether you have a genuine gap. If your telemetry is spread across four tools owned by three teams, you are shopping for something that sits above them, and native depth matters less than integration breadth.

Then ask what the correlation is actually trained on. A platform that has your dependency graph will outperform one inferring relationships from alert timing, and that difference shows up as false root causes rather than as a missing feature.

Finally, ask about the cost model before the capability. AIOps pricing tends to follow data volume, and the whole premise of the category is ingesting more telemetry than you did before. Several buyers have discovered the economics of that only after the correlation started working.

Forrester does not endorse any vendor named here, and tier placement should not be read as a recommendation to buy.