A few weeks ago, I was three thousand meters up a mountain, on a 14-degree grade, deep in what cyclists call the red zone: Blood pounding, lungs burning, nothing left to give but the next pedal stroke. And the whole time, I wasn't thinking about a release date. I wasn't thinking about a defect notice. I wasn't thinking about a security alert waiting for me back at sea level.
That's not because nothing was happening at Apica while I was gone. It's because my team had it handled. No calls. No panicked Slacks. Just people doing their jobs well enough that I could do mine, which, for two weeks, was surviving a climb.
I'm telling you this not because it's a cycling story. I'm telling you because it's the whole point of what we've just shipped.
The problem was never the people. It was the toil.
For years, the promise of observability tooling has been "more visibility." More dashboards, more alerts, more data. What nobody wanted to admit out loud is that all that visibility came with a tax: Someone had to configure it. Someone had to learn the YAML. Someone had to sit at 2 a.m. hand-writing a pipeline rule while a system quietly degraded around them.
That person is who Ascent 3.0 is for.
With this release, we've put artificial intelligence into the actual mechanics of the work, not as a chatbot bolted onto the side of the product, but woven through documentation, configuration, pipeline building, dashboard authoring, and fleet management. Someone on my team can now describe what they want in plain language, build me a pipeline that cuts data volume by 40%, and the system does the YAML-wrangling that used to eat their afternoon. That's not a party trick. That's giving someone their evening back.
Ownership, not obligation.
There's a philosophy underneath this that matters more to me than any single feature: We don't own your data, and we don't own your AI either.
Bring your own model, Anthropic's Claude, OpenAI, whichever certified LLM your compliance team has already blessed, and plug it into our open MCP server. We're not asking you to route your prompts through our infrastructure or trust us with information you're not allowed to share. If you're a bank, a hospital system, or a government agency where "we'll host your AI for you" is a nonstarter, this was built with you specifically in mind.
Maybe it's because we're a European company, and we've had to think harder about data sovereignty than some of our American counterparts have had to. Whatever the reason, it's not a marketing position for us. It's the same principle we've applied to your telemetry data since day one, now extended to the intelligence layer sitting on top of it.
The team that used to be stuck waiting.
If you've ever worked in observability, you know this feeling: Your platform engineering team is buried, your ticket is at the bottom of the queue, and meanwhile something in production is quietly falling over with no one watching it. You beg. You wait. Nothing moves, because deploying a data collector isn't their emergency, it's yours.
Ascent 3.0 hands that control back to the people who feel the pain when it's missing. Fleet management is now something your own team can direct with AI, independent of whoever else's priority list you'd otherwise be sitting on. You're no longer dependent on someone else's Terraform script or someone else's afternoon.
No, this isn't about replacing anyone.
I want to say this plainly, because I've spent enough of my career in rooms where it needed saying: Nothing in this release is designed to make your people disposable.
Every agentic action in Ascent 3.0 requires a human to approve it. Not because the technology can't act on its own — it can — but because we don't think it should, and neither do the enterprises buying this software. I've sold automation into markets where the pitch of "this replaces your headcount" was not just distasteful, it was a losing argument. It still is.
To start with, SysAdmins, SREs, DevOps, CSOs, ITOps – these are my people! I am first and foremost trying to help them. I know their jobs are hard, complex, and skilled – even with all the AI help in the world, enterprises still need them to own these roles. I also know they all have more than enough work to keep busy, even if AI takes away most of their 'toil'. And no LLM contains their years of development, and all the acquired 'tribal knowledge' they have of your specific infrastructure, your specific applications, your specific business, so they will always give you an edge.
The way I think about it: If you get rid of your people, you're getting rid of competitive advantage. What we've built is meant to let your strongest people do more of what only they can do, instead of burning their hours on repetitive configuration work a machine was always going to be better at. It will also help junior staff accomplish more simple tasks on their own, learning from examples as they go, without distracting those busy Seniors. Everyone gets more done, at their pace, with fewer interruptions. The flow must go on!
Proof, not a promise.
We've also leaned harder into synthetic monitoring in this release, and I'll admit, watching some of our competitors "discover" synthetics again this year has been a little vindicating. We never left. Connecting front-end digital experience to back-end tracing, catching the kind of AI failures that don't show up in a CPU graph, step-skipping, hallucination, quiet drift in a multi-agent workflow, that's been core to how we think about reliability from the start. We're not announcing a plan. We're already running it in production.
This is chapter one, not the whole book.
I don't want to oversell where we are. This release is a first big step toward intelligence in the pipeline, but it is not the last one. A bigger investment in Vanguard is already underway for later this year, and the agentic capabilities we've built into Flow and Fleet will keep expanding. So, we're not finished. In fact, with all I have on my roadmap, we're only just getting started.
Andi Mann is Chief Product & Technology Officer at Apica. Ascent 3.0 is available now.
