10/07/2026
For years, the working assumption in analytics was that you needed to know the code: SQL, Python, Java, whatever it took to coax an answer out of the data. Remzi Kurshumliu, Data Analytics Solutions Lead at Google EMEA, argued in this week's GIMI Think Tank that this assumption is fading, and that the harder problem has moved elsewhere. His half-joking line to colleagues is that the language to know now is English, since a clearly stated intent can lead an agent to the relevant reports, a prototype, and an explanation of what the data means.
But the part worth sitting with is what sits underneath. Traditional datasets were built for people reading charts and static measurements, while agents need explicit business relationships, conceptual definitions and pathways. He called the current BI setup a legacy pipeline: it extracts data, layers semantic views and produces insights for humans to interpret, but the warehouses beneath it lack the context and automated reasoning agents depend on. The numbers he shared show why this matters. 92% of organizations plan to increase AI investment, yet 70% struggle with the data unification and governance needed to succeed, and 90% of critical business data stays locked in unstructured formats.
On Monday morning, the useful question may not be which agent to buy, but whether your data can be understood by one.
The takeaway: agent-ready data comes before agent-led decisions.
๐๐ณ ๐๐ผ๐'๐ฑ ๐น๐ถ๐ธ๐ฒ ๐๐ผ ๐๐ฎ๐๐ฐ๐ต ๐๐ต๐ถ๐ ๐๐ฒ๐๐๐ถ๐ผ๐ป ๐ฎ๐ป๐ฑ ๐บ๐ฎ๐ป๐ ๐บ๐ผ๐ฟ๐ฒ, ๐ฏ๐ฒ๐ฐ๐ผ๐บ๐ฒ ๐ฎ ๐ฃ๐ฟ๐ผ๐ณ๐ฒ๐๐๐ถ๐ผ๐ป๐ฎ๐น ๐ ๐ฒ๐บ๐ฏ๐ฒ๐ฟ: https://lnkd.in/ggTwtBZ9
hashtag hashtag hashtag
