The race to build reliable artificial intelligence that can understand complex business data has taken a significant step forward, with a leading analytics firm launching what it claims is a definitive solution to the problem of inconsistent and untrustworthy AI-generated insights.
ThoughtSpot has announced Spotter Semantics, an “agentic semantic layer” designed to act as a crucial translation engine between raw, fragmented data stored in corporate systems and the AI agents increasingly tasked with analysing it. The company argues that without such a governed layer, organisations risk getting contradictory answers from their AI tools, undermining decision-making.
The Core Challenge: Context for Machines
At the heart of the issue, according to ThoughtSpot, is a lack of full context. When business users, data engineers, or AI agents ask questions of data in different ways, the answers must remain consistent. Francois Lopitaux, SVP of Product Management at ThoughtSpot, stated that a robust semantic layer has always been part of the company’s DNA, developed from the ground up to support natural language search rather than retrofitting AI onto static dashboards.
“Critically, this deterministic approach relies on our patented search tokens, not text-to-SQL powered by LLMs, which is why we can guarantee the most consistent, trustworthy insights on the market,” Lopitaux said.
The system is built on an AI-native foundation, using a specialised query generation engine and AI-powered indexing to convert natural language into complex SQL. It references knowledge graphs that integrate business logic, security rules, and metric definitions into a machine-readable format. This architecture is designed to mitigate AI hallucinations and misinterpretations by providing agents with a governed, context-aware framework.
Governance, Performance, and an Open Framework
As companies grant data access to thousands of users and AI agents, governance becomes paramount. Spotter Semantics centralises a “single version of truth” in a governed Metrics Catalog to prevent metric drift, which can cause leaders to question data reliability. Analysts can create custom metrics and formulas via a visual interface, while data engineers prepare underlying data with SQL and developers manage deployments via APIs.
To address performance and cost, a feature called Aggregate Awareness intelligently routes queries to either detailed or pre-aggregated tables based on the question, aiming to ensure fast response times while reducing compute costs.
Emphasising interoperability, ThoughtSpot is a founding member of the Open Semantic Interchange (OSI) standard, creating a vendor-neutral abstraction layer between cloud data warehouses and AI tools. This framework is intended to keep business logic portable, allowing integration with existing models in platforms like Snowflake and Databricks. Furthermore, the ThoughtSpot Model Context Protocol (MCP) server allows businesses to connect their semantic layer directly to any AI agent or large language model.
Measured Adoption and Customer Endorsement
The company reports growing traction for its agentic approach. By the end of fiscal 2025, the ThoughtSpot platform surged 133% year on year, and over 64% of all customers—from Fortune 500 companies to startups—actively use its Spotter AI as their primary AI analyst.
Manbir Paul of beauty giant Sephora, speaking on *The Data Chief* podcast, highlighted the transformative impact. He noted that enabling data consumers with ThoughtSpot helped the company understand business concepts behind their data and enrich their semantic layers based on how users explore information, which he said “has driven a lot of value for us.”
Looking ahead, ThoughtSpot’s commitment to what it terms the “agentic era” is driving further innovation for Spotter Semantics, with future capabilities set to include writeback for actionable analytics and Federated AI Search.
