๐—ช๐—ต๐˜† ๐—ฌ๐—ผ๐˜‚๐—ฟ "๐—”๐—ด๐—ฒ๐—ป๐˜๐—ถ๐—ฐ ๐—•๐—œ" ๐—ฆ๐˜๐—ฟ๐—ฎ๐˜๐—ฒ๐—ด๐˜† ๐—ถ๐˜€ ๐—•๐˜‚๐—ถ๐—น๐˜ ๐—ผ๐—ป ๐—ฎ ๐— ๐—ฒ๐—น๐˜๐—ถ๐—ป๐—ด ๐—œ๐—ฐ๐—ฒ๐—ฏ๐—ฒ๐—ฟ๐—ด

On February 3, 2026, Snowflake launched ๐—ฆ๐—ฒ๐—บ๐—ฎ๐—ป๐˜๐—ถ๐—ฐ ๐—ฉ๐—ถ๐—ฒ๐˜„ ๐—”๐˜‚๐˜๐—ผ๐—ฝ๐—ถ๐—น๐—ผ๐˜. This tool automates the creation of semantic views to ensure AI agents operate on trusted business metrics rather than guessing definitions from raw tables. Simultaneously, research from ๐—š๐—ฎ๐—ฟ๐˜๐—ป๐—ฒ๐—ฟ and ๐—ง๐—ฟ๐˜‚๐˜€๐˜ ๐—œ๐—ป๐˜€๐—ถ๐—ด๐—ต๐˜๐˜€ predicts that through 2026, organizations will abandon ๐Ÿฒ๐Ÿฌ% ๐—ผ๐—ณ ๐—”๐—œ ๐—ฝ๐—ฟ๐—ผ๐—ท๐—ฒ๐—ฐ๐˜๐˜€ due to a lack of "AI-ready" data foundations.

The industry is finally admitting that general-purpose agents "rarely survive real workloads" because they lack the deterministic grounding of a semantic layer.

๐—ง๐—ต๐—ฒ ๐—ฆ๐—ฐ๐—ต๐—ผ๐—น๐—ฎ๐—ฟ ๐—ฉ๐—ถ๐—ฒ๐˜„: ๐—•๐—ฟ๐—ถ๐—ฑ๐—ด๐—ถ๐—ป๐—ด ๐˜๐—ต๐—ฒ "๐—ฆ๐—ผ๐—ฐ๐—ถ๐—ผ-๐—ง๐—ฒ๐—ฐ๐—ต๐—ป๐—ถ๐—ฐ๐—ฎ๐—น ๐—š๐—ฎ๐—ฝ"
The failure rate of these AI projects (60%โ€“80%) validates the ๐—ฆ๐—ผ๐—ฐ๐—ถ๐—ผ-๐—ง๐—ฒ๐—ฐ๐—ต๐—ป๐—ถ๐—ฐ๐—ฎ๐—น ๐—š๐—ฎ๐—ฝ. While the "Technical" capability (LLMs/Agents) is accelerating, the "Social" or organizational capability (Governance/Metadata) is stagnant.

This idea is well known in academic literature. Organizations are attempting to deploy "Autonomous" systems (High Maturity) on "Siloed" data estates (Low Maturity). Without a ๐—ฆ๐—ฒ๐—บ๐—ฎ๐—ป๐˜๐—ถ๐—ฐ ๐—Ÿ๐—ฎ๐˜†๐—ฒ๐—ฟ acting as a "translation layer," the agent cannot reconcile the gap between technical schemas and business meaning, leading to the "Amnesiac Analyst" syndrome.

๐—ง๐—ต๐—ฒ ๐—ฃ๐—ฟ๐—ฎ๐—ฐ๐˜๐—ถ๐˜๐—ถ๐—ผ๐—ป๐—ฒ๐—ฟ ๐—ฉ๐—ถ๐—ฒ๐˜„: ๐—Ÿ๐—ฒ๐˜€๐˜€๐—ผ๐—ป๐˜€ ๐—ณ๐—ฟ๐—ผ๐—บ ๐˜๐—ต๐—ฒ ๐—›๐—ฒ๐—ฎ๐—น๐˜๐—ต ๐—ฆ๐˜†๐˜€๐˜๐—ฒ๐—บ ๐—™๐—ฟ๐—ผ๐—ป๐˜๐—น๐—ถ๐—ป๐—ฒ๐˜€
During my tenure as Interim Director, I saw exactly why "Cortex Code" or "Semantic Autopilots" are necessary but not sufficient. You don't solve project failure with better SQL; you solve it with ๐——๐—ฎ๐˜๐—ฎ ๐—Ÿ๐—ถ๐˜๐—ฒ๐—ฟ๐—ฎ๐—ฐ๐˜† ๐—ฅ๐—ผ๐˜‚๐—ป๐—ฑ๐˜€.

If a clinician asks an agent for "Length of Stay," and the agent hits a raw table without a governed semantic definition, it might pull "Midnight Census" while the clinician needs "Discharge-to-Discharge." In a healthcare environment, that delta isn't just a dashboard error; itโ€™s an ๐—ผ๐—ฝ๐—ฒ๐—ฟ๐—ฎ๐˜๐—ถ๐—ผ๐—ป๐—ฎ๐—น ๐—ฟ๐—ถ๐˜€๐—ธ. The "Semantic Layer" isn't just a technical feature; it is the ๐—œ๐—ป๐˜๐—ฒ๐—ฟ๐—ป๐—ฎ๐—น ๐—”๐—ฃ๐—œ ๐—–๐—ผ๐—ป๐˜๐—ฟ๐—ฎ๐—ฐ๐˜ that allows an agent to speak "Metric" without hallucinating.

Originally published on LinkedIn

https://www.linkedin.com/posts/malikalamin_%F0%9D%97%AA%F0%9D%97%B5%F0%9D%98%86-%F0%9D%97%AC%F0%9D%97%BC%F0%9D%98%82%F0%9D%97%BF-%F0%9D%97%94%F0%9D%97%B4%F0%9D%97%B2%F0%9D%97%BB%F0%9D%98%81%F0%9D%97%B6%F0%9D%97%B0-%F0%9D%97%95%F0%9D%97%9C-%F0%9D%97%A6%F0%9D%98%81%F0%9D%97%BF%F0%9D%97%AE%F0%9D%98%81%F0%9D%97%B2%F0%9D%97%B4%F0%9D%98%86-activity-7425282343977959426-Jx47?
utm_source=share&utm_medium=member_desktop&rcm=ACoAAAGjt7sBL8uj9adPfrG1EfHYraXT1G5wf0s

Previous
Previous

๐—ง๐—ต๐—ฒ ๐—ฆ๐—ฒ๐—บ๐—ฎ๐—ป๐˜๐—ถ๐—ฐ ๐—Ÿ๐—ฎ๐˜†๐—ฒ๐—ฟ: ๐—ง๐—ต๐—ฒ ๐—ฅ๐—ผ๐˜€๐—ฒ๐˜๐˜๐—ฎ ๐—ฆ๐˜๐—ผ๐—ป๐—ฒ ๐—ณ๐—ผ๐—ฟ ๐—”๐—œ ๐—”๐—ด๐—ฒ๐—ป๐˜๐˜€.

Next
Next

๐—ช๐—ต๐˜† ๐—ฌ๐—ผ๐˜‚๐—ฟ ๐——๐—ฎ๐˜๐—ฎ ๐—ช๐—ฎ๐—ฟ๐—ฒ๐—ต๐—ผ๐˜‚๐˜€๐—ฒ ๐—œ๐˜€๐—ปโ€™๐˜ ๐—ฅ๐—ฒ๐—ฎ๐—ฑ๐˜† ๐—ณ๐—ผ๐—ฟ ๐—”๐—œ ๐—”๐—ด๐—ฒ๐—ป๐˜๐˜€.