OptroFrom bottlenecks to self-service: How Optro accelerated team productivity with Grafana Cloud and Grafana Assistant
At Optro, the company behind the industry-leading software GRC system of action, observability data is essential. But until recently, that data was effectively locked away, reserved to a small group of internal experts. The result: longer investigations, more frequent escalations, and highly skilled engineers spending time answering questions instead of solving high-impact problems.
But with Grafana Cloud and Grafana Assistant, that all changed. Optro was able to transform its observability into a self-service capability—reducing escalations, accelerating resolution times, and enabling more than 60 engineers and support staff to independently investigate issues.
The result was a measurable increase in productivity across the organization, turning observability data into faster, more confident business decisions.
A team of experts in a world that needed more of them
Peter Wagenet leads the Developer Experience (DX) team at Optro. His team has a broad mandate—covering CI/CD infrastructure, developer productivity, and AI tooling adoption. Puneet Kandhari leads the Resilience Engineering team, responsible for Resilience, Performance, AI infrastructure, and data pipelines.
For Optro’s FIRE org, a group of roughly 50 engineers, the challenge was straightforward in theory but difficult in practice: across that group, only around 10 people had deep Grafana expertise. Everyone else at Optro—developers, QE/SDET engineers, Tier 1 and 2 support—had access to observability data, but faced a steep learning curve to understand and utilize it.
“There’s maybe three people on my team, and 10 people on the Infra team who are Grafana experts,” Puneet Kandhari, who leads the Resilience Engineering team, explains. “And yet, we have 100 people on the Product, Engineering, and Support teams who need access to that data to better do their jobs.”
The implication was a familiar one in engineering organizations: the people who needed observability insights had to go through the people who knew how to find them.
Building the onramp
The turning point came with Grafana Assistant and, specifically, “quick start” prompts —customizable preset queries that give users a structured starting point before they ever type a question.
Kandhari’s team built two prompts that changed the support workflow: a namespace investigation prompt and a synthetic monitoring alert prompt. Once deployed, support engineers could run structured investigations independently. They didn’t need to know PromQL, ping the Resilience team, or wait in a Slack queue.
“Once we had quick start prompts, we were able to just set those up for everybody, and that is why the adoption was high,” Kandhari says.
The support workflow shifted noticeably. Previously, when one of Optro’s customers reported latency issues, the path looked like this: customer → CSM → Tier 2 support → Slack escalation → engineering investigation. Now, Tier 2 runs the prompt themselves, cross-references the results with an internal knowledge base of previous resolution patterns, and decides whether to escalate or just fix it.
“We’ve trained them that instead of posting in Slack and asking for help, run this prompt,” Kandhari explains. “Look at the data. And then they have a solution, and it’s just a question of: here’s what we saw last time, here’s what we’re supposed to do. And we’re like, go for it.”
Time to resolution decreased. Escalations to engineering dropped. And Kandhari’s team could focus on the issues that actually needed expert judgment.
CI observability: building what GitHub couldn’t
On the Developer Experience (DX) side, where Senior Staff Engineer Peter Wagenet leads the team, their use case was different, but the problem was the same. He knew the data existed. He just needed a way to interrogate it.
Optro runs a high-throughput CI pipeline with hundreds of GitHub Actions workflows running continuously. One metric they particularly needed (total time a commit spends in the merge queue) isn’t something GitHub surfaces natively. It requires pulling run times from multiple workflows for the same commit and synthesizing them into a single view.
Wagenet used Grafana Assistant to build that query through an iterative back-and-forth conversation.
“If you look at the query, you’re like, wow, this is kind of crazy,” he says. “I can see how maybe I could have figured it out myself. But having to do that, it would have taken me a lot longer.”
The result was a historical performance dashboard that, for the first time, gave Wagenet’s team measurable, longitudinal CI data. It proved immediately useful: the team brought CI queue times from roughly 25 minutes down to 18 minutes – and the dashboard gave them the evidence to make the case for further investment.
“I’ve taken numbers directly from that dashboard into leadership discussions,” Wagenet says.
The dashboard didn’t just tell the story. It made the case.
Observability beyond the experts
One of the less expected outcomes has been how Assistant has helped to spread usage across the organization. Wagenet even cited an executive who is using it. “I don’t think [he] was going in there and making his own dashboards prior to Assistant,” he says. “There are definitely people who are now doing things they just probably wouldn’t have otherwise.”
For Wagenet, this represents something more significant than efficiency gains: “It’s not just that it makes things faster—it lets you do other things you just wouldn’t have done.”
Even cost management has become a collaborative effort. Kandhari describes using Grafana Assistant to identify expensive dashboards through the fair-use drilldown view, then optimizing those queries—a proactive workflow the team built organically after gaining better visibility into their own usage.
What’s next: Claude meets Grafana Cloud
Optro is now working on a deeper integration that will extend Assistant access even further. The team recently gained access to the Assistant CLI, and an engineer is building a Claude Code skill that connects Claude’s coding agent directly to Assistant.
The vision: when an engineer gets a customer bug or PR to review, their Claude instance can query Grafana directly, pulling trace data, profiles, and metrics on the relevant services without switching contexts or opening a browser.
“Every engineer who has Claude will also get the skill that gives them Grafana,” Kandhari says. “When an engineer gets a report for a customer issue, they’re just gonna use Claude to talk to the Assistant to do what they need.”
For the support side, the next goal is to trigger custom quick start prompts directly from Slack, with a specific customer namespace and time window. That would close the final loop: investigation, insight, and resolution, all without leaving the tools engineers already live in.
The longer-term vision is more ambitious: automated pre-release performance analysis as a standard part of the development cycle. Every release gets a K6 load test. Every run generates traces and profiles. Assistant runs a standard investigation and shares results with the developers whose code is in the release before anything ships.
“It’s basically incident management, but before the incidents hit the production line,” Wagenet says. “And it’s not just incidents—it’s cost incidents, latency incidents, and actual bug triages.”
