Participant: Datamole

Abstract: This is an experience report on building software with autonomous agents on a real project. We had to rewrite a frontend of a mid-size application, and improve its UX along the way. Our usual approach would be mockups, then a high-fidelity prototype, then implementation. Instead, we used an agentic workflow to build a working prototype in the stack we’d use for the real implementation.

Autonomous agents all share one problem: drift. Left on its own, an agent wanders from what you ask, and early mistakes compound the longer it runs. Most of the work is in controlling that. The agent needs the right context to work from, clear rules for how it should operate, and checkpoints that catch drift before it spreads.

We’ll go through the workflow we built, the challenges we hit while making it reliable, and what we learned running it.

Speaker: Petr Svoboda is a Solution Architect at Datamole. He worked as a full-stack engineer for 10+ years and holds a master’s degree in web and software engineering from the Faculty of Information Technology of the Czech Technical University in Prague. Most of his career has gone into replacing manual, spreadsheet-driven work in industry with software that helps teams make better decisions. He’s now applying the same instinct to software development itself, working toward building applications with autonomous agents.

Speaker: Petr Nevyhoštěný is a Lead Engineer at Datamole. He holds a master’s degree in theoretical computer science from the Faculty of Information Technology of Czech Technical University in Prague and has applied his expertise across a wide range of problems over the last 10 years, from custom numerical solver implementation through advanced testing approaches to web systems architecture. Beyond his standard duties, he now drives the agentic engineering capability of Datamole’s Application Development department: establishing shared workflows and onboarding the team of eight developers. He enjoys teaching and writing long-form technical articles, both on his personal blog and for Datamole’s engineering publication on Medium.