What is one specific practice or element that is essential for building a reliable semantic layer to support AI-driven analytics in organizations, and why?
Most answers emphasize the need for trusted, well-governed, and clearly defined data and metadata as foundational to building a reliable semantic layer for AI-driven analytics.
12 answers
Answers


Matt Ausman
Chief Information Officer · Zebra Technologies
Data Governance
Matt Ausman says data governance is essential, pointing to Zebra's federated model with data owners per function, because running AI on unvalidated data or shifting definitions will give bad results.
The answer starts at 16:17.
Technovation with Peter High (CIO, CTO, CDO, CXO Interviews) · Listen on Apple Podcasts

Shayde Christian
Chief Data & Analytics Officer · Cloudera
Technical and Business Metadata
Shayde Christian says technical and business metadata was very important in creating a semantic layer, and that his organization made a lot of strides in that area while continuing to build toward a true semantic layer.
The answer starts at 31:07.
Chain of Thought | AI Agents, Infrastructure & Engineering · Listen on Apple Podcasts
Mike Aronson
Vice President, Global Talent Acquisition · Johnson Controls
Central Ontology
Mike Aronson says the ontology part is where defining information once and reusing it across models lives, and describes comparing semantic views to find common pieces across multiple models.
The answer starts at 56:43.
Explicit Measures Podcast · Listen on Apple Podcasts

Martin Casado
General Partner · Andreessen Horowitz
Open Table Formats
The passage does not show Martin Casado giving this answer; the recommendation to use open table formats, specifically Iceberg, is made by George Fraser.
The Analytics Engineering Roundup · The Analytics Engineering Roundup · Read

Anahita Tafvizi
Chief Data & AI Officer · Snowflake
Unified Metric Definitions
Anahita Tafvizi says organizations need unified definitions for their metrics, because without that foundation AI can just scale your chaos across the company.
New York Stock Exchange · Watch on YouTube

Jess Ramos⚡️
Founder & Content Creator (Tech, Data, & AI) · Big Data Energy Analytics⚡️
Trusted Data
Jess Ramos says having the right trusted data or trusted intelligence is going to be critical for AI to work, because an AI agent given the wrong information will fail, hallucinate and do the wrong thing.
Jess Ramos · Watch on YouTube
Mike Ferguson
Owner, CEO and Head of Research · Intelligent Business Strategies
Data Concept Models
Mike Ferguson says he is seeing organizations go back to their roots by getting data concept models in place, then building and curating data around them to achieve a systematic, repeatable approach.
The answer starts at 12:57.
It's About Data · Listen on Apple Podcasts

Francis Rose
Host & Co-founder · Fed Gov Today
Understanding Data Landscape
Francis Rose says the most important thing is understanding the landscape of your data, including its full scope, where it comes from, and who should access it, because the models and AI tools are only as good as that data.
Fed Gov Today · Watch on YouTube

Gaurav Pathak
Senior Vice President Product Management · Salesforce
Single Source of Truth
Gaurav Pathak says organizations need a single source of truth for metadata and data, framing it as essential for getting agentic AI right and worth a big investment in data.
Super Data Science: ML & AI Podcast with Jon Krohn · Watch on YouTube

Tristan Handy
CEO & Founder · Dbt Labs
Clean Gold Layer
Tristan Handy says a clean gold layer is essential, meaning high-quality data with no duplication of tables or columns, and eliminating that ambiguity is really important because agentic analytics plugs directly into the gold layer.
The answer starts at 19:15.
DataFramed · Listen on Apple Podcasts

Sam Pierson
Chief Technology Officer · Qlik
Quantifiable Data Reliability
Sam Pierson says having a data foundation that is quantifiably reliable and high quality is the first thing his team found necessary, along with adding context on top of the data to improve the accuracy of LLM answers.
Data Faces Podcast · Data Faces Podcast · Listen at 8:12

Muralidhar Vemulapalli
Chief Enterprise Architect/ Acting CTO · Arkansas Blue Cross And Blue Shield
Enterprise Semantic Layer
Muralidhar Vemulapalli says defining an enterprise semantic layer is the most important element, because machines, unlike developers, do not know what column names mean.
CDO Magazine Podcast Series · CDO Magazine Podcast Series · Listen at 4:11

