Quick answer: At Data Expo in Utrecht, nobody seemed worried about whether AI works anymore. The question was: will it work here? Inside this data landscape. With these legacy systems. With these processes. With data that may be "good enough for a dashboard" but looks rather less good once an AI agent starts making decisions with it. The second theme: companies want more capability with less ceremony – fewer handoffs, fewer layers, more people who understand the problem and can act on it. AI is getting easier. Enterprise reality isn’t.
The shallow lesson: pick a better model
The model was rarely the problem. The environment it lands in is.
The real lesson: the hard part isn’t getting AI to do something impressive
It’s getting it to do something useful inside a real company, and that takes people who can see the problem end to end, not another layer of handoffs.
The sobering part
Data that fed reports for years can fail the moment an agent acts on it. Errors that were cosmetic on a dashboard become operational in a decision. Different continents, different conversations, same uncomfortable bit after the hype.
Full disclosure: moving Data & AI from PoC to production is what Azati does, so we’re partial to "environment first." See our engineering work.
FAQ
Adapted from a LinkedIn post by Eugene Volkov, Commercial Director at Azati Software – read the original, written after Data Expo 2026 in Utrecht, with a nod to an AI TechWorld post by Alex Tsimashenka.
