At DrupalCon Rotterdam, our team immersed themselves in the latest developments across the open-source ecosystem. CCO Ulla Koho and Head of Technology Tomi Mikola were joined by Senior Developers and long-time Drupalists Mario Vercellotti and Hannes Kirsman. Beyond taking in the keynotes and technical sessions, Ulla and Tomi stepped onto the stage. They delivered two presentations: on Monday they presented “Who owns the answer”, together with Kaisa Liukkonen from University of Helsinki. On Wednesday Ulla and Tomi spoke at the main event DrupalCon, with a session called “Nobody asked for a CMS: from content management to answer delivery in the age of AI.”
How we experienced DrupalCon — the perception vs. reality gap
A recurring theme throughout the event, echoed by Drupal founder Dries Buytaert in his Driesnote, is that public perception of Drupal lags about four years behind reality. Too many assumptions still reflect what the platform was doing in the immediate post-COVID era. In truth, modern Drupal has leapt far ahead of many other open-source alternatives, retaining its core architectural strengths while evolving to meet today’s digital demands.
”We have to be loud about it, seriously. Drupal is like the platform we knew, but on steroids. It fits perfectly with the modern need for digital sovereignty, open-source flexibility, and high security—while delivering editorial and front-end experiences that no longer fall short of any competitor.”
— Ulla Koho, CCO at Mearra
Key takeaways from Rotterdam
- Revolutionised editorial experience: The evolution of Drupal Canvas proved how effortlessly the platform now handles intricate, multi-brand, multi-language setups with strict editorial control. The case study presented by Boyan Borisov (Jakala) and Anthony Bojocan (Mars) demonstrated how modern Drupal balances global scale with precise local governance.
- Drupal as an AI & integration hub: Drupal’s structured content architecture makes it an ideal central integrator connecting web services, external databases, and AI tools. Your AI, your choice, no lock-in, as said in the Driesnote. On top of that, you get AI guardrails, one source of truth and every interaction logged.
- Unconstrained front-end flexibility: Headless and modern front-end capabilities have removed historical design and user experience limitations.
- Core strengths amplified: Open source, security, and digital sovereignty remain central pillars, giving enterprise and public sector organisations complete ownership over their digital infrastructure.
As Dries noted during his keynote: no one else will shout about these advancements for us. The community and agencies alike need to make these capabilities clear to the broader industry.
What we presented — “Nobody asked for a CMS”
In our session with Kaisa Liukkonen from the University of Helsinki, we addressed a simple truth that digital teams often overlook: users don’t care where information lives, nor do they ask for a CMS. They just want an answer they can trust.
”Nobody ever cared where the information lived. They wanted an answer.” — Ulla Koho
”Adding the AI wasn’t the hard part. Everything else was.” — Tomi Mikola
Proof point: HelsinkiUni Help
We presented real-world evidence from HelsinkiUni Help, an AI search solution built for the University of Helsinki that has been in active production for 10 months:
- 70,000+ questions answered across 119 countries.
- 1,000+ questions per day during peak term-start periods.
- 33,633 distinct questions recorded, with 80.8% asked only once, proving that real user queries cannot be predicted by traditional navigation menus alone.
Lessons learnt: why content governance precedes AI
Building an AI answer engine is not primarily a technical challenge of plugging in a large language model. The hard work lies in organisational clarity, content ownership, and governance:
- Answers need owners before search engines do: Content across intranets, study guides, and public sites must have designated owners who take responsibility for accuracy, timeliness, and context across channels.
- Access & security safeguards: When handling protected intranet content, services must decide whether to exclude restricted sources, require login up front, or show synthesised answers while keeping underlying source pages protected behind authentication.
- Observability & “morning coffee digests”: Automated monitoring digests tracking pipeline health ensure content imports run smoothly overnight, giving product owners visible proof that indexed data is complete and reliable.
- Keeping humans in the loop: AI models can misapply rules out of context (e.g. applying a specific grant deadline to general admission requirements). Continuous learning loops, where unanswered or misrouted questions are flagged, turn query gaps into actionable content improvements.
3 questions for your organisation
If your institution or enterprise is evaluating AI search or content strategy, consider these three starting points from our presentation:
- What keeps the problem alive?
- Are we solving the problem, or just moving it?
- Are we in control of the decisions, or just in the loop?
Interested in seeing how governed content management and AI answer delivery work together? Ulla is keen to discuss in more detail, so send her a message or share your contact details and we will get back to you.
P.S. If you want to know what level your organisation’s AI maturity is at, check this quick AI maturity test out!
