letsboot organises the Agentic Engineering Days Zurich 2026, Switzerland’s AI Coding & Agents Conference. It takes place on 20 November 2026 at the Marriott Hotel Zurich – a seven-minute walk from Zurich main station. We expect 300–400+ attendees from engineering teams that already run AI-assisted coding and agents in production, or are about to. Full conference details are on www.agenticdays.com.

This page lists the current programme: 38 sessions across four tracks – talks, spotlight talks, lightning talks, hands-on sessions and a panel discussion. Every entry links straight to its detail page with the full abstract.

Talks

• “YOU MUST NOT ssh!” and Other Stories from Building Our SRE Agent
  David Pech (Wrike)
  Hard lessons from months of running an SRE agent in production — framework and model matter less than you think.
• Agents Speak Protocol: Why Standards Are the Real Infrastructure of AI
  Stefano Maestri (IBM) · Alessio Soldano (IBM)
  Every computing revolution ended in standards — why protocols, not applications, are the real AI infrastructure.
• AI & Agility: Is Spec-Driven Development Bringing Back Waterfall? And How to Move Beyond It
  Baptiste Macé (Arolla)
  Spec-driven development promises a lot — but is it waterfall returning, and how do you move past it?
• Beyond Prompts & Models: Evolving Enterprise Agentic Systems
  Manisha De (Swisscom)
  Not models and frameworks, but the recurring architecture decisions and trade-offs that shape enterprise agentic systems.
• Building Authorization for Agentic RAG Systems
  Sohan Maheshwar (AuthZed)
  Agentic RAG in the enterprise needs fine-grained authorization — the missing piece, and how to build it.
• Caveman and the Empty Room: Designing Agents for an Audience of Zero
  Julius Brussee (Caveman)
  A Saturday-night skill made agents talk like cavemen; three weeks later tens of thousands of developers were using it.
• Deploying Agents at Scale: What Enterprise Adoption Actually Looks Like
  Michael Forrester (Accenture)
  From inside enterprise rollouts: where deployments and research agree, and where they contradict each other.
• Human-in-the-Loop Is You: A Guide to Working With Multiple Coding Agents
  Oleksandra Bovkun (Databricks)
  Working with Claude Code, Codex and Cursor at once without pasting the same bug report into three windows.
• Let’s build our own Claude Code
  Max Gfeller (Sutro)
  Coding agents feel like magic, but the core ideas are approachable — we build one from scratch.
• Migration or migraine: What no one tells you about using LLMs
  Warren Parad (Authress)
  The eternal backlog task: upgrade every framework without wrecking coverage or production. Perfect for AI. Or so we thought.
• OCR at Swiss Post: The Case for Small Language Models
  Huriel Reichel (Swiss Post) · Oliver Hofmann (Swiss Post)
  Reaching for a frontier LLM by default costs money, latency and privacy — small models cover most bounded tasks.
• Reachy mini: Giving a body to AI
  Andrés Marafioti (Hugging Face)
  LLMs reason and talk but stay behind screens; an open, developer-friendly robot gives AI a body.
• Something New from NVIDIA
  Ricardo Silveira Cabral (NVIDIA)
  Ricardo Silveira Cabral brings something new from NVIDIA; topic and abstract follow closer to the event.
• The Agentic Lab Factory: How I Stopped Writing Labs and Started Reviewing Them
  Raphaël Pinson (Isovalent)
  Getting new SMEs to write labs was always too costly — what does AI change about that equation?
• The next frontier - agentic collaborative environments.
  Krzysztof Cieślak (GitHub Next)
  Agentic engineering has been a solo story so far; what’s missing is planning, understanding and team alignment.
• Watching Agents for Trouble
  Natalie Pistunovich (HTW Berlin)
  Once an agent can deploy or send mail, the question shifts from answer quality to how much damage goes unnoticed.
• We Rolled Out GitHub Copilot… How Do We Prove It Helps?
  Yuliya Khadasevich (Xebia) · Liuba Gonta (Xebia)
  Licences paid, usage high — but how do you actually prove GitHub Copilot helps?
• When Software Stops Being Deterministic: Rethinking Engineering Practices for AI-Native Systems
  Susanna Wong (Typeform)
  Software was always deterministic. What happens to tests, reviews and CI when it stops being so?
• You Can’t Ctrl-Z a Robot: Coding Agents in Embedded Systems
  Syrian Hadad (adesso Schweiz) · Alexander Büchel (adesso Schweiz)
  In the terminal feedback is free. In the physical world there are no error messages and no undo.
• Your Biggest New Customer Can’t See Pixels
  Tim Damen (ABN AMRO)
  Accessibility was always the right thing to do. The argument gets easier when your next customer isn’t human.

Spotlight Talks

The Spotlight track belongs to the conference sponsors and partners – their tools, products and new work.

• 21 million exceptions a day: getting Bedrock and AgentCore right
  Sven Beck (BLP Digital) · Marco Kuendig (copebit)
  How BLP Digital resolves 21M+ exceptions a day for 550+ enterprise customers, and how copebit runs the AWS foundation.
• Agentic AI at Migros Online: Autonomous Engineering Agents and Conversational Shopping
  Sven Scheffel (Migros Online) · Gabriel Dinant (Migros Online)
  Agents at both ends of e-commerce: delivery work inside engineering, and customers shopping through AI assistants.
• Close the Lid: Coding with AI Agents from Your Phone
  Sebastian Witalec (Namespace)
  Agents can do real work — but the machine running them doesn’t have to be the laptop under your arm.
• From Copilot to the Dark Factory
  Pini Reznik (re:cinq)
  AI makes developers faster, not organisations. Review queues and platform bottlenecks just move the constraint.
• FSI Transformation at Scale: Sovereignty, Compliance & Production Realities
  Sergiu Petean (Atlassian)
  Every financial institution wants to be AI-native. How realistic is that under sovereignty, compliance and production constraints?
• Specs first, code later (or how I build software in the AI-DLC era)
  Alessandra Pasini (AWS)
  Specification first, code later: the three-phase AI-DLC workflow, starting at inception.
• Tie a Metric to Every Release
  Matt Elwell (Flagsmith)
  Every release behind a flag and tied to a metric — rolled out and watched live on stage.

Lightning Talks

• A Practical Guide to Fast and Private LLMs
  Lena Fuhrimann (bespinian)
  Practical tips for running your own LLM privately: architecture, optimisation, hardware and software stack.
• From a RAG Demo to a RAG System
  Richa Dinesh Tamhane (amnis)
  Almost every agent that looks something up runs RAG; when retrieval is wrong, the agent is confidently wrong.
• Running Secure Agents at Scale
  Carlos Sanchez (Adobe) · Alex Carol (Adobe)
  Multi-tenancy, noisy neighbours, tool access, data boundaries — secure agents need a deliberate execution model.
• Securing MCP in the Enterprise: OAuth, Token Exchange, and Real-World CI/CD Integration
  Valentin Delaye (ELCA)
  Integrating MCP into an on-prem CI/CD toolchain (SonarQube, Jenkins, Backstage …) with Keycloak as identity provider.
• Welcome Session
  Jonas Felix (Co-Founder Agentic Days / letsboot.ch) · Barbara Felix (Co-Founder Agentic Days / letsboot.ch)
  Barbara and Jonas Felix open the conference.

Hands-On Sessions

75–90 minutes of building rather than watching: bring a laptop and follow along. Seats go by arrival, no registration.

• ADK Crash Course - From Beginner To Expert
  Abdel Sghiouar (Google Cloud)
  Build autonomous agents with the Agent Development Kit — planning, reasoning and tool use, not chatbots.
• Agent on the Wire: Looking Inside an AI Coding Agent
  Vincent G (letsboot)
  What actually happens between a task and working software — an AI coding agent seen from the inside.
• Build Your Own Production-Ready Agentic Apps
  Kevin Dubois (IBM) · Dimitris Andreadis (Red Hat / IBM)
  Evolve a simple LLM client into a production-minded agentic app with Quarkus and LangChain4j.
• Hands-on Spec-Driven Development
  Simon Martinelli (martinelli.ch)
  Develop software systematically with Spec-Driven Development and the AI Unified Process.
• Just Build — Orchestrating your Agents with Codex
  Abbas Tutcuoglu (OpenAI) · Arsenii Gavrilenko (OpenAI)
  A hands-on lab from OpenAI: orchestrate agents with Codex, step by step on your own laptop.

Panel Discussion

• Can We Trust Autonomous Agents? (Panel Discussion)
  Maxim Salnikov (Microsoft) · Jose Luis Latorre Millas (Swiss Life) · Serena Sensini (TheRedCode) · Markus Leutwyler (VisionPilot)
  How much autonomy can agents get in production? Guardrails, evals, observability and human oversight. Moderated by Maxim Salnikov.

Tickets and speakers

• Tickets for Agentic Engineering Days Zurich 2026
• All conference speakers