The beer wasn’t the only thing flowing in Munich last week. From June 8–12, 2026, the Westin Grand Munich was taken over by the network automation community as the Network Automation Forum hosted AutoCon5.
I’ve been to a lot of industry conferences over the years. NANOG, Cisco Live, AWS re:Invent. You name it. But AutoCon has carved out a uniquely honest space in that landscape. No vendor CEO keynote. No pay-to-play sessions. Just 700+ engineers sitting in a room together, willing to admit what isn’t working and collaborate on how we make progress together. It’s my kind of event.
NAF does one conference a year in the United States and one in Europe, and the city changes each time. I’ve followed this community for a while, and getting to attend in Munich for the third European edition felt like a milestone worth being there for. I always learn something at AutoCon and wanted to share a little bit of those learnings here.
What a Week it Was
The event kicked off Monday and Tuesday with two full days of pre-conference workshops. Sixteen sessions spanning everything from Python basics and config management with Nornir and NAPALM, to advanced topics like building a chained agentic AI system to cut MTTR and standing up a modern observability stack. The workshops required a separate ticket and several sold out, a sign that practitioners showed up wanting to get their hands dirty, not just collect slide decks. I was not able to attend the workshops this time around but I heard great things from those that did. I highly recommend checking them out at a future AutoCon. The proctors do a great job with these.
The main conference ran Wednesday through Friday, and the energy in the room was noticeably different from a typical vendor-driven event. When Michael Bushong from Nokia stepped up to deliver the opening keynote on “The Cognitive Biases Behind Automation’s Failures and Future Successes,” he set the tone immediately. The argument? Our industry has poured billions of dollars into automation tooling, and most networks remain largely un-automated. The problem isn’t the technology. We all know the protocols work, the frameworks are mature. The problem is us. Loss aversion, survivorship bias, and misframed business cases have quietly killed more automation initiatives than any protocol gap ever did. It was a great way to open the conference, and his point was exactly right.
The Talks That Stuck
Wednesday afternoon featured a variety of practitioner-led sessions that reinforced Bushong’s thesis from the ground up. DE-CIX’s Lucas Immanuel Nickel walked through two production-ready delivery patterns his NetOps team ships via GitLab pipelines in under ten minutes. Sony PlayStation’s Tim Sando made the case that ops-led automation shouldn’t wait for a top-down program that rarely moves fast enough. Start with one script, build trust, and grow from there. Deutsche Bahn’s Leo Fleskes shared something that really resonated with me: two years of data work had to happen before any automation was possible. A VPN rollout that used to take 70 minutes across four teams now takes 25 across one. That’s the return on doing the unglamorous work first.
Thursday went deep on both the technical and organizational dimensions of automation. Swisscom’s session on reaching fully automated backbone networks (10,000 devices, SRv6, NSO, zero manual intervention) was the kind of talk that makes you simultaneously inspired and humbled. Eric Chou’s session on navigating the LLM landscape for network engineers was a practical and much-needed map of an AI ecosystem that has gotten very crowded, very fast.
The NAF Framework Track ran in parallel on Thursday and I heard great feedback from those who attended. I wasn’t able to catch those sessions in person, but I’m looking forward to the videos once they’re posted. Claudia de Luna’s walk through the NAF Solution Wizard on a live use case was apparently a standout, a reminder that the unglamorous work of reference architecture and shared vocabulary is what makes this community actually move forward together.
Building the Business Case for Agentic Tool-Chaining
I had the privilege of presenting in the Leadership Track on Thursday afternoon with a talk titled “The ROI of Network Autonomy: Building the Business Case for Agentic Tool-Chaining.” The Leadership Track ran in parallel with the general session, and I’ll admit there’s a certain pressure that comes with competing for attention against a room full of technical content. But I firmly believe the only way network automation moves forward is if we, as engineers, learn to speak the language of our business. We have to learn to articulate the ROI of the systems we are building.
The core argument: automation in networking isn’t a research project anymore, but the people writing the checks still aren’t sure it’s worth the investment. Tool-chaining, connecting LLM-driven agents to real operational tools like your Source of Truth, your monitoring stack, and your ticketing system, is where the real ROI lives. Not in any single tool in isolation. But making that case to a VP or a CFO requires a different kind of fluency. It’s a sociotechnical problem that requires the organization to be ready for it, not just the engineers.
I walked through how to frame the business case: what metrics actually matter, how to quantify the cost of man hours spent on tasks that automation can handle, and how to present risk-adjusted returns in a language that lands with decision-makers. The room was engaged. I saw people taking pictures of slides, which is always a good sign. I’m looking forward to the survey results so I can keep improving, but if you want to dig into the business case further, find me on LinkedIn or Bluesky.
Final Thoughts
AutoCon 5 made something clear: the network automation community has grown up. The conversations in Munich were more grounded than ever. A lot of people are now doing network automation, not just talking about it. AI for NetOps and autonomous networking is becoming more real, albeit with guardrails needed. The hallway track was buzzing all week with genuine peer-to-peer problem solving, not vendor pitches dressed up as thought leadership.
These are just a few of the many excellent talks delivered at AutoCon5. If you missed it, the recordings will be worth your time. And if you want to put the next one on your calendar, AutoCon6 is coming to Tucson, AZ, November 16-20, so block it now. This is the conference our community has needed for a long time, and it keeps getting better.
Prost, Munich. See you next time.