Alexander Pilz: Fifteen years inside an EU agency: what makes a software partnership last

This is part of the business track at the Plone Conference 2026 in Maastricht, The Netherlands. After the keynote, this lays the ground.

Durability creates sovereignty. But durability in a partnership should not make switching hard.

If you look at platforms, you hope they live longer than the initial contract. We have been supporting a platform for 15 years. The procurement was all about price, capability, compliance, and exit terms. No one at that time thought about data retention, about if the contractor would actually care about the project, and would push back in case decisions were made that in their view could damage the project. But over 15 years that is needed.

We at Syslab already knew EU-OSHA in 2005, doing a website for them. In 2011 a European launch was done for a worker risk assessment platform, adapted from a similar Dutch platform, but now with open code. We made lots of progress since then.

Time is not the asset. What survives through time is.

• Institutional context: why constraints and exceptions exist.
• Technical memory: why were choices made
• Mutual understanding: communicate before a problem becomes a crisis
• Practiced coordination: who acts, decides, reviews, and is informed.

A communications officer wrote as review of us that they liked our "critical perspective to avoid embarking in grandiose but useless projects." For example we could have spent lots of effort into turning the website into an app, but we didn't, as it would not really give much extra value compared to the effort put into it.

Sometimes, the "captain" on the customer side leaves to a different job. That has happened, and then we as contractor could carry the organisation through a transition period before a new captain was found, as we know the project so well.

Exit must be possible, but replacement is not costless.

We thought about where AI fits inside this project. For a long time we have wanted to create training videos, in multiple languages. With AI this becomes possible. Content structure: finding patterns in vast amounts of content, for example risks that happen in almost all cases. AI makes sense there. But we won't set up a chatbot, giving automated advice to end users. Especially for this risk assessment tool you would very quickly run into compliance issues.

A durability test for procurement: ask for evidence, not declarations of partnership values.

• Show us a relationship that survived changes in staff and technology?
• Describe a client request you opposed.
• How does your model reward simplification?
• Where does the code and data stay when the partner leaves?

Exit must remain real. Disagrement must remain possible. Staying must keep earning its value. That is when durability created sovereignty.