Leading digital banking platform accelerates incident response and democratizes observability with Grafana Assistant

A leading digital banking platform serves hundreds of financial institutions across the U.S., where uptime, performance, and rapid incident response are essential. As the company shifted from its previous observability vendor to Grafana Cloud, it began exploring Grafana Assistant – the AI-powered agent in Grafana Cloud – as a way to help teams troubleshoot faster and gain deeper ownership of their systems.

For the Director of Service Operations, the appeal was immediate. Instead of relying strictly on static dashboards, their teams could ask questions and Assistant would “respond like an engineer” rather than a generic AI bot.

Since adopting Grafana Assistant and Assistant Investigations – an SRE agent built directly into Assistant – the company has reduced mean time to resolution (MTTR) in some instances by up to ~60%, expanded observability access across engineering roles, and accelerated its cultural shift toward stronger DevOps ownership.

Bringing observability to everyone while managing a complex stack

The Director of Service Operations oversees five teams that support the company’s digital banking platform, including a 24×7 network operations center (NOC), service management, and a newer observability function. Their environment spans both legacy and modern services, a complex hybrid setup that often required expert knowledge to troubleshoot effectively. Newer engineers and support team members frequently needed help refining questions, understanding dependencies, or navigating the telemetry available to them.

In parallel, the company was pushing for stronger developer ownership. As services onboarded into Grafana Cloud, developers needed a more intuitive way to explore and interpret their data without waiting on the observability team.

Overall, the company needed a way to make observability more accessible, reduce reliance on tribal knowledge, and scale incident response across a wide range of experience levels.

AI-guided insights, grounded investigations, and custom onboarding

Grafana Assistant quickly became central to NOC workflows. Analysts now consult Assistant when an issue surfaces, asking what the AI agent sees in the live telemetry.

“We’ve already helped reduce our MTTR for incidents – in some cases, from 60 minutes down to 25 because instead of manually having to look for answers, Grafana Assistant Investigations gave us the answer right away,” the Director of Service Operations said. “In the middle of an incident, that type of time difference is gold.”

The benefits extend to seasoned engineers as well. For complex incidents that previously took “three hours to figure out,” Assistant Investigations now surfaces relevant components and metrics “really darn quick.”

Because of this, the company plans to make Assistant Investigations part of their SEV-creation workflow to help analyze and declare severity levels for incidents. The director said they are “very much looking to switch” SEV creation to query Grafana Assistant Investigations because they see it “providing better answers for how to solve problems.”

The company also uses Assistant to outline an RCA report, since the agent automatically correlates metrics from before, during, and after an incident. While it doesn’t create a complete report on its own, Assistant still significantly reduces the time and complexity involved in that process, which is a huge win for the team, the director said.

An AI assistant tuned to understand the platform’s architecture

To ensure Grafana Assistant could reason effectively across the company’s hybrid architecture, the Director of Service Operations trained it with a detailed, 100-page internal document. “We’ve had to teach Assistant how our platform works,” they explained.

The director compared the process to onboarding a new team member – one who learns extremely quickly. After providing historical context and architectural details, the director found that Assistant answers questions as well as one of the company’s “best engineers,” especially when asked how systems connect or why certain behaviors occur.

“One of the most powerful differentiators of Grafana Assistant Investigations, compared to other AI tools, is that it actually has native access to our telemetry data to learn from,” the director explained. “Other tools could help answer questions, but having access to all of our signals, in real time, means we don’t lose time setting context or teaching the AI each time we have a question to answer.”

Self-service observability for developers

As more development teams onboard Grafana Cloud, Grafana Assistant has become a force multiplier for enablement. Developers who once relied on observability experts now start by asking Assistant for help building dashboards or alerts, the director said.

“Developers come and ask my observability team a question and the first reply is always, ‘Have you tried asking Grafana Assistant?’” Often, within minutes, the developers then have “a large portion of their data visualized” using Assistant without any formal training.

This shift supports the company’s broader DevOps strategy. Developers not only visualize their data faster, but also use Assistant to identify gaps in monitoring. The director said developers now ask the AI assistant where gaps exist, and “it’s going to spot every gap and make so many suggestions,” reinforcing stronger product ownership – a key goal the company has identified for 2026.

A communication style teams trust

One of the most impactful outcomes has been Assistant’s ability to support both high-level engineers and junior support staff. The director observed that the AI agent “can basically talk to high-level engineers one-to-one,” while also helping less technical team members refine their questions until it can guide them effectively.

The director also tuned Assistant’s tone to be more concise and direct – similar to how engineers talk when troubleshooting – and feedback has been overwhelmingly positive. After making it “a tad more terse,” teams found the responses clearer, more actionable, and less prone to hallucination. “People are loving Assistant now,” they said.

Looking ahead

The company plans to deepen its use of Grafana Cloud IRM, integrate Assistant Investigations more tightly into SEV workflows, and continue improving developer enablement through AI-driven insights and automated RCA support.

In the meantime, teams are actively testing the boundaries of Assistant’s capabilities. In one experiment, the Director of Service Operations dropped a full Slack room’s 100-message investigation into Grafana Assistant. It parsed the entire discussion, summarized what happened, and even offered to build dashboards and continue the investigation.

“We don’t even know where the limits are yet,” the director said. “And we’re going to keep pushing.”