A new global survey suggests artificial intelligence is pushing the engineers who keep systems running to rewrite their playbooks as it moves from pilot projects into everyday business operations. Dynatrace’s “The State of SRE and Platform Engineering 2026” report, based on responses from 919 senior IT leaders across 16 countries, paints a picture of two mature disciplines under fresh pressure.
Two disciplines, one shared headache
Site reliability engineering (SRE) and platform engineering have been around for years: SRE since 2003, platform engineering since 2018 and both are now standard practice at large enterprises. But the report says adoption is no longer the issue. The real challenge is scale: as AI workloads move into production, they behave unpredictably, drift over time, and can fail without obvious warning signs.
Watching the machines
The most common AI use case among SREs, cited by 58%, is monitoring AI systems themselves, checking model performance, accuracy, and data security. Platform engineers, meanwhile, are focused on giving developers access to AI tools like copilots and chatbots (55%). In short, both groups are using AI cautiously, prioritising visibility before handing over more control.
That caution shows up in the numbers. While AI is delivering real benefits, results are trailing expectations almost everywhere, most notably in cost savings (45% actual versus 55% expected) and faster incident resolution (46% versus 53%).
Old tools, new strain
Traditional reliability tools are buckling under AI’s unpredictable nature. Service-level objectives (SLOs) are used by 89% of teams, but half say too many data sources get in the way. Quality gates, automated checkpoints in software pipelines, are well established too. Yet, the report argues they can’t fully vet AI on their own, since AI output is probabilistic rather than fixed. Continuous, real-time monitoring for “model drift” is becoming the new safety net.
Platform teams battle integration, not adoption
For platform engineers, internal developer platforms (IDPs) are now the norm: 89% of platform teams have one, and 60% say it’s broadly used. But growing pains persist: 37% cite trouble integrating existing tools and systems, the single biggest complaint, ahead of maintaining standards and managing security or compliance.
Governance goes AI-native
Encouragingly, security and compliance appear to be built in rather than bolted on. Roughly two-thirds of SRE and platform teams embed security policies directly into their platforms as code, and a similar share automate compliance checks. The report frames this as good news: organisations that already treat governance as code are best positioned to extend those habits to AI systems.
Observability: the underused superpower
Perhaps the report’s central argument is that observability the practice of monitoring systems for performance and health needs to evolve into what Dynatrace calls a “control plane” for AI. Nearly all SREs (97%) already use some form of observability, and 65% use dedicated tools. Yet only 40% of platform engineers have it fully embedded across all deployments, leaving real potential on the table.
The bottom line
The report’s message is one of measured optimism: SRE and platform engineering have reached operational maturity, but AI is a new kind of workload that demands new kinds of oversight. Teams that connect observability, automation and AI into one trusted system, the report argues, will be the ones equipped to run AI safely and reliably at scale.