Grafana is offering its AI assistant for free to open source and on-prem users — though on stage at its Barcelona user conference this week, CEO Raj Dutt joked they shouldn't use it too much.

The observability vendor also took the wraps off Grafana 13 and a re-architected version of its Loki log aggregator, while dropping strong hints about expansion into broader business analytics and workflow beyond its observability tools.

Grafana 13 includes curated prebuilt dashboards tailored to specific data sources, templates around methodologies like DORA, and a generally available dynamic dashboards feature. Other additions include a redesigned query editor, a refreshed saved queries interface, and giving users the ability to create panels directly from saved queries. Git Sync, now generally available, lets ops teams manage "observability as code" within GitOps workflows.

But you can't have a keynote these days without mentioning agents: Dutt announced the Grafana Assistant - previously cloud-only - is coming to on-prem and open source users. They'll need to connect to a Grafana Cloud account to handle the LLM connection, though observability data stays on-prem. The assistant can power automations in Slack or Teams, or it can be accessed via API or CLI.

Grafana Cloud also gains AI Observability, promising real-time visibility into AI agent behavior. Principal engineer Mat Ryer described it as a "10,000-foot view" of where time, tokens and "therefore dollars" are going, with the ability to drill down into agent activity — including spotting policy violations or data exposure.

Dutt said: "I'm honestly really excited [about it] but also quite afraid… that we're offering Grafana assistant to the 99 percent. It's not just going to be limited to the 1 percent of Grafana users [who pay]."

"I was talking to our chief financial officer about this, and he was kind of worried. So I'm excited – please use it, but maybe not too much."

• Google claims to have all the answers for enterprise AI agent sprawl
• Headless 360: Salesforce's latest pitch to let AI do the dev work
• If an AI agent screws up while running your business, there's nobody to sue
• AIOps is so powerful, vendors are building tools to clean up after agents break your infrastructure

Grafana presumably hopes to make up for that by monetizing the use of its platform beyond observability. "We want to take everything that we're doing in observability, and bring it to the rest of the business over time. So that means looking at things not just like requests, but maybe revenue, not just latency, but maybe LTV."

He said Grafana saw a convergence of "business data and observability data, especially as firms become more and more software led."

Product lead Jen Vila added the team was experimenting with features in open source Grafana, before deciding whether to formally launch features in Grafana Cloud.

"I would say that our main hypothesis right now is there is a growing desire to look again at these business analytics, metrics alongside your observability data. And so that's maybe like a wedge we could start with."

On the infrastructure side, the Loki rework brings Kafka-based ingestion and a redesigned query engine and scheduler. The key win: reducing redundant writes. Engineering manager Poyzan Taneli said the previous iteration could "end up storing the same log 2.3 times… based on object storage. This is a lot when you operate at the petabyte scale." ®