AI Reality Check 2026: a full hall and a conversation about AI reality, not promises

The ANIMAS conference "AI Reality Check 2026" took place at the Riga Motor Museum on September 30. The interest was so great that we increased the initial number of 100 seats, and additional spots were also filled.

The goal of the conference was to discuss what happens in the implementation of AI after ideas, presentations, and initial experiments: what works, what doesn't, what we have learned, and where the risks begin.

What we heard from the stage

• Gatis Romanovskis (ANIMAS) opened the conference with a thought that characterized the entire day: AI is like fire, with which we all must learn to handle.
• Betija Muižniece (ANIMAS) in her presentation "My most productive colleague is artificial intelligence" showed how AI works in ANIMAS's daily operations, where simple automation is sufficient and how to achieve more with the same people.
• Gatis Ozols (VARAM) spoke about the journey from AI strategy to real application in public administration.
• Jānis Grants (DPD Latvia) demonstrated that there is also immense room for digitalization, automation, and robotization in logistics: "What cannot be digitized can be automated."
• Lilita Beķere (Numeri) shared her experience of implementing AI simultaneously in three Baltic countries and why it is a matter of management and change, not an IT project.
• Jānis Ozols (Estimo.lv) talked about how data and AI are changing the real estate industry: faster analysis, quicker identification of risks, and a more complete understanding of properties.
• Kristaps Cīrulis (SHIFT2, Shipyard AI) explained why so many AI initiatives do not meet expectations and why good technology alone is not enough.
• Guna Puce (Artificial Intelligence Center) discussed the collaboration between the state and entrepreneurs: "From parallel realities to a common solution."
• Jekaterina Macuka (Data State Inspectorate) spoke about what we trust AI with, where our data goes, and how to maintain control over personal data.
• Elviss Strazdiņš (bool) presented a live demo showing what happens when a language model's safety and refusal mechanisms are removed.

The panel discussion included Gatis Romanovskis, Jekaterina Macuka, Guna Puce, Kristaps Cīrulis, and Sanda Dejus (Radio SWH), who brought in the human perspective – what these changes mean for someone who simply encounters AI in their daily life. The conference was moderated by Valdis Melderis.

Thank you to the speakers and everyone who came, listened, asked questions, and discussed. After this day, it is clear: one such event will not be enough.