The Snowball Has No Brakes: Governance, Humans, and AI at DevDays Europe 2026
Earlier this month, members of the 542 Digital team made the trip to Vilnius, Lithuania for DevDays Europe 2026, a four-day software development conference co-located with DevOps Pro Europe and CyberWiseCon. One ticket, three conferences, and more opinions about AI than you could shake a large language model at.
The venue deserves a mention: Multikino, a working cinema, complete with an IMAX screen in one room and an audience Q&A system that let attendees post questions and vote in live polls without interrupting the speaker. Small details, but they matter when you're absorbing a lot of content across four days.
The programme covered a broad spread, from psychological wellbeing to MCP server governance, with AI woven through almost everything. What emerged was a consistent and somewhat uncomfortable theme: the technology is moving considerably faster than the systems designed to keep it in check, whether human, organisational, or regulatory.
The Psychology of the Engineer in the Age of AI
Povilas Godliauskas opened with something refreshing: an evidence-based talk backed by documented datasets rather than opinion. Burnout and chronic stress; all of these make engineers less effective and more difficult to work with, with contractors and freelancers affected disproportionately.
The AI findings were sharp. Moderate AI use correlates with genuine performance gains and reduced workload. High AI use eliminates those gains entirely. There's an optimal zone, and the industry currently has the subtlety of a sledgehammer when it comes to locating it.
He also made a point worth pinning above every product roadmap: organisational stress (chasing funding rounds, hitting IPO targets) leaks directly into design decisions. We've seen this with engagement-driven social platforms built to be addictive rather than useful. We're starting to see it with AI products. Don't build a solution until you've found the problem.
Automation: Quietly Building the Cage
Janis Janovskis, Lead DevOps Architect here at 542 Digital and later a panellist, delivered a useful corrective to the prevailing enthusiasm. Automated systems are often a source of workplace stress rather than a relief, accumulating that quality gradually until someone notices the cage has been locked from the inside.
More pointedly: automation can remove a user's ability to intervene when things go wrong. In software, that's an incident. In safety-critical systems, the consequences are considerably more serious. The question is always whether a human can meaningfully take back control when it doesn't work.
MCP Governance: Don't Let the AI Near the Database
Horatio Gonzalez's session on MCP server governance was practical and well-structured. MCP (Model Context Protocol) is the emerging standard by which AI agents communicate with external tools and services. It's powerful, increasingly widespread, and, if implemented carelessly, a reasonably efficient way to hand an AI the keys to your infrastructure.
The headline advice: put an API between your AI and your database. This sounds obvious until you consider how many teams are not doing it. Beyond that, the session covered prompt injection risks, authentication requirements for every MCP call, and a useful reframe on evaluation. "Does it work?" is not the right question. "Does it work correctly?" is better: does the model choose the right tool, with the correct arguments, in the correct order? MCP requires substantially more testing than a traditional API.
AI, DevOps, and a Revealing Audience Poll
Alex Shershebnev's talk on AI in DevOps workflows sat in productive tension with Godliauskas's findings. The practical guidance was sound: GitHub stars are not a security indicator; auto-scan everything before ingestion; use different models to review code for higher assurance. Treat AI tooling with the same scrutiny you'd apply to any third-party dependency, which is to say: considerably more than most teams currently apply.
The session opened with a live poll on AI use at work. The proportion of the audience not using AI at all, primarily due to compliance and governance constraints, with a notable contingent from banking, was larger than expected. Governance is lagging. Which brings us to the panel....
Panel: DevOps in the Era of AI
The panel, featuring Costa Tsaousis (Netdata), Olle Pridiuksson (QA Tech), Vlatko Ivanovski (Anglo American), and our own Janis Janovskis, surfaced the gap between smaller experimental organisations and larger enterprises focused on governance and spend control.
The employment question came up: what happens to junior developers? The broad consensus was that required knowledge is shifting. "Developer" is giving way to "Engineer": someone who understands how services connect, rather than someone who has memorised the syntax of a particular language. Our own experience at 542 Digital reflects this; AI has allowed the team to work competently across languages and frameworks outside our core speciality, in ways that simply weren't practical before.
The Actual Takeaway
The AI snowball is large, accelerating, and largely indifferent to the governance structures in its path. The banking professionals in the room not using AI aren't necessarily being overly cautious; their organisations are responding to real regulatory requirements written for a different technological moment. Governance is lagging, not slightly but structurally, and that gap is the real story of where the industry is right now.
Janis Janovskis, who joined the panel after his own session, put it well:
AI is no longer being treated as magic; it is being evaluated as another powerful engineering tool that must be understood, governed, and integrated properly. The tooling is changing rapidly, but the need for strong engineering thinking remains central.
The consistent message from Vilnius was that humans need to stay in the loop, not as a nice-to-have, but as a design requirement. The carbon-based parts of the system still matter, probably more than the current pace of AI adoption would suggest.
DevDays Europe is worth the trip. Good speakers, genuine breadth, excellent chairs, and the rare experience of watching an IMAX-scale slide deck about burnout statistics. We'll be back.