Reflections on AI Week, and the future of solving problems with observability and AI
Thank you for spending AI Week with us. We’re thrilled by the reaction and we all enjoyed replying to your questions. Thanks for engaging.
Some of my favorite quotes from LinkedIn and Reddit include:
“Grafana’s latest announcements aren’t just 'AI features.' They signal a shift in what observability platforms are becoming.”
“I had Claude migrate and update and validate dozens of hand-rolled dashboards… Worked astonishingly well… 95% done in 30m with the MCP and Claude.”
Good vibes
While AI Week is coming to a close, the story doesn't end here. Just like you, we're in the middle of a major transition in how AI and observability work together to solve problems quickly.
We can’t do it without you. So please try our tools in your own teams, and let us know what we got right, as well as where we need to keep iterating. We know lots of people already feel left behind by the rate of change, so we’re making it our mission to catch you up, filter out all the noise, and bring you the very best in agentic operations.
That's why I'd like to use this last post for the week to look back at how we got here and look ahead at what's yet to come—for Grafana Labs' AI tooling, for the software development and management industry, and most importantly, for users like you.
Our story
At Grafana Labs, we've always tried to innovate at the cutting edge of observability to solve real problems brought to us by our internal and external users. But when it comes to LLMs, we were a little late to the party. To be completely honest, we were skeptical. Starting in 2024, we began seeing so many examples of AI features being crammed in where they didn’t belong. And even when the use case made sense, the tools seemed incapable of executing the actual task.
We realized things had changed when we started seeing LLMs running by themselves in loops; when agents were born.
There is little doubt that agents have changed the way we build software. You don’t have to look far to find examples of people who have been supercharged by Claude, Codex, or Copilot, even if they aren’t loop-maxing, mob-vibing, or pitting agent against agent in adversarial arenas.
We are building more and shipping more thanks to agents. Not only that, we’re also working in entirely new ways. What do you do while your agents are thinking? Some engineers are working on multiple projects at the same time. And I mean that literally; while one agent cooks, they are prompting another.
The rate of change has dramatically increased, and change, as we all know, is where trouble starts. Each engineer is now producing more operational demand than ever before and this shows no signs of slowing down. Naturally, at Grafana Labs, we feel like it falls to us to help you and your teams adapt and thrive in this new world.
The AI team at Grafana Labs has grown significantly over the past year, but we remain dedicated to the experimentation and innovation that is already helping thousands of users today. We will continue to hold ourselves to the highest standard: we will never ship AI slop tools, or AI for the sake of it. To do this properly, we need your help. Your feedback, good and bad, helps ensure we’re solving real problems and it helps us continue to test our assumptions.
A recap of AI Week
Here’s a refresher of how the week went.
Now generally available:
- Agent Observability: No more black-box agents. Watch what they do, evaluate their output, and spot problems early
Also in preview and ready to play with:
Alongside the big reveals, the team also wrote about the thinking behind it all. There's the case for baking observability in from the very start, and how giving your coding agents access to your telemetry keeps you confident in what you ship. There's also how to automate the operational grind, and guidance on how to build trust in your agents, born from the challenge of scaling Grafana Assistant itself. You can find all of them over on the Grafana blog.
We also lifted the lid on how some of this actually gets built. There are behind-the-scenes stories on gcx (our command line tool for humans and agents) and Assistant Search, and a look at how we build at Grafana Labs.Those can all be found on our engineering blog, Unprompted.
So that’s a wrap for our first AI Week. Let us know how you think it went, what you’d like more of, and where we can improve if we do this again.
On a personal note, I hope you see the dedication and creative spirit that the team is pouring into our new tools. We’re having a lot of fun building in this exciting space, and we can’t wait to hear from you after you have had time to kick the tires.
What’s next?
So, where do we go from here? Things are continuing to change out there at an exciting and daunting pace. We’re starting to trust AI on more meaningful tasks, and we’re seeing places it can run reliably as part of our workflows. We’re already impressed by some of the skills and MCP combinations people have set up for Assistant and Assistant Investigations, integrating it even deeper into their systems and teams.
At Grafana Labs, we’re shaping a future where AI removes the daily operational grind, empowering engineers to focus on what’s next. As telemetry insights become universally accessible across the organization, we anticipate a shift toward more data-driven requests and deeper, enlightened discussions, ultimately allowing us to make better decisions.
Auto-improvement loops can help optimize code based on your telemetry (which could be anything from infra metrics to business data), run A/B tests, and roll out the winning combination. Bug reports from users can be pre-investigated before a human has even seen it, enriching the support experience and speeding up resolution. Automated agents can carry out janitorial duties and generate intelligent reports that land in the right place at the right time.
It's a lot to adapt to, but we plan to be there with you every step of the way. Thank you again for hanging out with us during AI Week, and please join us at ObservabilityCON 2026 in San Francisco this October for an update on all-things AI from Grafana Labs.
Grafana Assistant is the easiest way to get started with metrics, logs, traces, dashboards, and more in Grafana Cloud. We have a generous forever-free tier and plans for every use case. Sign up for free now!
