Most enterprises will hand root cause analysis to AI agents within two years

For decades, software engineers have been the eyes and ears of enterprises, combing through voluminous logs, observability tools, consoles, browser tabs, and other application data. The goal? To find and fix pattern discrepancies between code aberrations and known steady and safe code states.

But those human-driven observations are often slow and require large numbers of engineers to manage ever-expanding workloads and identify problems and threats. The engineers must manually form hypotheses, investigate the data and incidents, and then determine the best remediation plan.

Today, a more sophisticated approach exists. 

IT observability has been gaining powerful new capabilities. According to Elastic’s 2026 report, The Landscape of Observability, 85% of organizations use some form of GenAI for observability. Within two years, that number is expected to reach 98%. Since the introduction and growth of generative AI (GenAI) and agentic AI, GenAI’s role is expanding, helping enterprises shift from reactive observability to a more proactive and adaptive stance. Using GenAI, enterprises can now create AI agents that autonomously monitor their IT systems, interpret data and telemetry, and understand what went wrong, enabling them to improve systems and performance.

As this transition from largely manual observability interventions continues, intriguing and time-saving new use cases are emerging, involving GenAI investigations of observability data. These advancements are aimed at core team members, including site reliability engineers (SREs) and IT operations workers, but they are also providing valuable new opportunities for other business teams across enterprises, such as DevOps, cybersecurity, compliance, and line-of-business employees.

For example, product managers can use agentic AI and GenAI tools to run A/B tests on new releases to obtain regional and segmentation-level conversion data. And finance teams can use GenAI assistants to review service level agreements (SLAs) with external service providers to quickly determine whether SLAs are being met.

The impact of GenAI for observability data and how AI agents can help

By moving to agent-managed observability tools, enterprises can now transform how SREs, security teams, developers, and others work by freeing them from the burden of constantly managing and securing their infrastructure, Thaddeus Walsh, Principal Solutions Architect at Elastic, tells The New Stack.

“I see this as attention economics. If developers didn’t have to worry about the underlying components that are necessary to ensure an application runs properly and could just focus on delivering functional capability, that would be perfect for them.”

“I see this as attention economics,” says Walsh. “If developers didn’t have to worry about the underlying components that are necessary to ensure an application runs properly and could just focus on delivering functional capability, that would be perfect for them.”

By incorporating AI agents to evaluate and use the same telemetry signals, including logs, traces, and metrics, that are evaluated by human IT workers, the agents can autonomously change the configuration states of applications, said Walsh. “That would be an ideal universe, where no human ever has to touch it.”

So, why is observability important to enterprise IT teams? 

With effective observability tools and generative AI, teams can evaluate, monitor, and improve the performance of distributed IT systems much more effectively than using traditional manual monitoring methods. This gives organizations deeper visibility and critical insights across the many data sources in cloud-native environments.

These tools also benefit root cause analysis (RCA) investigations, in which lines of code are evaluated to identify problems so they can be traced and corrected, says Walsh. “RCA investigations go deeper than operational investigations. An operational investigation might focus on how to return to a steady state. An RCA is about needing to understand what caused a problem.”

How GenAI and AI assistants are typically used by SREs and IT operations

Much of today’s AI-assisted investigation work is still done on the operational side of enterprises, says Walsh, as technicians work to understand the current state of the situation and then determine the fastest path to remediation.

“An SRE may be responding to an incident where some system is impacted, and they need to familiarize themselves with the context,” he says. “They also look at how that brokenness is expressed via the raw signal data and what is the most likely next step in the investigation. They look to find the most reasonable route to pursue validation for their hypothesis.”

Watch: Beyond the SRE: Democratizing observability data with GenAI

One of the biggest historical challenges of RCA investigations is that, to declare causality, you have to be able to identify the dependencies among the components of the system or application. Unfortunately, most of the related documentation is insufficient, which largely makes root cause analysis difficult in many contexts, Walsh explains.

Insufficient documentation is a key area where GenAI can boost observability investigations and improve root cause analysis for SREs and IT operations personnel, says Walsh.

Can organizational users outside of IT also benefit from observability data access?

Absolutely. Non-traditional users, such as salespeople, finance teams, compliance personnel, and other non-developers, can gain a broad range of benefits, including new opportunities to integrate the tools into their own investigations.

All the information captured and evaluated by the observability platform, including all the telemetry from logs, metrics, and traces, can give rich new access to illuminating customer and business data that can open new avenues for creativity, revenue, and better data analysis by a wider set of constituents within a company, says Walsh.

“The first barrier for users is access to the data, and the second piece is that the users must have the skills and knowledge to be able to interpret the raw signal data from the observability platform into a shape that is meaningful to that stakeholder,” he says. “It is a desire for data-driven decision making.”

Here’s what employees outside of IT can do using AI-driven observability

The value is enormous, says Walsh. Observability is a domain that defines data capture and analysis of system-generated telemetry data, all of which can help enterprises better tell their stories and make broader business decisions.

For example, logs and GenAI can help transform customer experiences by providing granular, real-time visibility and early warnings of system issues that could cause customers to leave a company’s website and head to other vendors. Logs can reveal subtle latency increases, intermittent errors, or degraded services before they escalate into widespread outages, and those signals can be mapped directly to Service Level Indicators (SLIs), such as response time or error rate, for resolution.

Logs are also important in providing revenue protection for companies because every performance issue, failed transaction, or downtime event carries financial risk for their bottom line. By using GenAI-powered, log-driven insights, companies can identify which issues affect revenue-generating services, high-value customers, or critical workflows, and then implement adjustments and fixes. 

Logs today are no longer just a tool for debugging. They are a source of truth, a driver of decisions, and increasingly, a competitive advantage for organizations.

How can organizations democratize their AI-driven observability to increase its value?

The power of agentic AI and AI agents is making a significant impact on enterprise observability across the industry today, but the changes are still underway, says Walsh.

“We’re in the midst of an important inflection point, because most organizations are still relying on human-driven processes to perform their GenAI observability and RCA investigations. In the next 24 months, it’s very likely that the majority of enterprises will not have human-led investigations. They will transition to agent-led investigations, with observability and data access across all of their systems.”

“In the next 24 months, it’s very likely that the majority of enterprises will not have human-led investigations. They will transition to agent-led investigations.”

This democratization of using AI-driven observability and processes across enterprises and their employees will be possible as companies reorganize their existing observability ecosystems to remove fragmentation and siloed systems, says Walsh.

“There are a lot of organizations that are tiptoeing with AI, but they have no automated orchestration, and humans have to approve every step. That’s a huge amount of friction,” he says. 

Want to learn more? Watch the “Beyond the SRE: Democratizing Observability Data with GenAI” webinar now.