AtScale's Semantic Layer Doubles AI Accuracy in Business Queries…

Event summary

AtScale's research shows its semantic layer doubled AI accuracy on business questions across seven models, with gains ranging from 1.7x to 2.4x.
The study used BIRD-Interact benchmark with 600 tasks across 22 databases, focusing on 410 business questions.
Google's Gemini 3.1 Pro saw a 345% lift in follow-up question accuracy with AtScale's semantic layer.
AtScale announced a new suite, AtScale for AI, integrating semantic modeling, question clarification, and accuracy evaluation.

The big picture

AtScale's findings highlight a critical gap in enterprise AI adoption: the need for consistent, governed business definitions to ensure AI reliability. As businesses increasingly rely on AI for financial and operational workflows, the ability to double AI accuracy through semantic layers could accelerate the integration of AI into core business processes. The research underscores the strategic importance of bridging the gap between probabilistic AI models and deterministic business metrics.

What we're watching

Adoption Pace: The pace at which enterprises will integrate semantic layers into their AI workflows to ensure consistent, auditable answers.
Model Performance: Whether open-weight models like Kimi K2.5 can sustain performance close to frontier models with the help of semantic layers.
Enterprise Trust: How the framework addressing prompt disambiguation, semantic resolution, dirty data handling, and complex metric calculation will impact enterprise trust in AI.