Business intelligence platform provider ThoughtSpot Inc., today announced what it says is a major expansion of the artificial intelligence capabilities it debuted last year that integrate human training into the query and analysis process.

The “human-in-the-loop” feedback control in ThoughtSpot Sage includes a patent-pending technology that enables human operators to customize and improve query prompts by providing associations between business terms and the data model. Customers can view and edit all queries, as well as user-generated feedback to deliver more accurate results.

“The reality is foundational models are good for language understanding but not for analytics,” said Chief Development Officer Sumeet Arora. “We are solving the problem of leveraging AI for what it is good at while providing trust in analytics.”

LLMs are good at understanding human expression and generating basic SQL but were never designed to formulate complex queries, he said.

“They’re not trained for handling multiple tables, columns, complex joins, filters and date operators,” he said. “The code generation benchmarks have been on very simple data sets that don’t reflect the real world.”

Verifiable answers

In the new version of Sage, “every analytical answer is user verifiable,” Arora said. “It’s retrieval-automated generation plus and LLM plus relational search for 100% accurate analytics.” Administrators can control who is allowed to train the model.

The company is also adding a feature it calls Ask Sage that gives users the ability to ask follow-up questions in natural language as they interact with data visualizations.

Data scientists and engineers can use the AI Assist feature ThoughtSpot acquired with its purchase of Mode Analytics Inc. last summer, to create, refine, understand, troubleshoot and optimize more technical queries.

AI features are also being put to work on Liveboards, the collection of related visualizations that comprise a business intelligence dashboard. The feature looks for changes in attributes and key performance indicators based on learned user activity and sends them as notifications. The KPI Monitor feature alerts users to changes in natural language via AI-generated change analysis delivered over the ThoughtSpot mobile app.

Embedding options

The company Is also expanding options for users who want to embed ThoughtSpot features in their own applications. ThoughtSpot Sage Embed lets users add ThoughtSpot Sage to any web or mobile application via an application program interface.

“This is one of the most exciting things we’re seeing,” Arora said. “People want a conversational interface, especially on mobile apps.” He said the embedded AI assistant lets users ask questions such as “Which store in the region has the highest sales?” or “Are we on track to hit our environmental goals?” without using a query language.

For developers, the new AskDocs feature lets developers ask coding questions in natural language and receive generative AI-assisted instructions and code from both ThoughtSpot and developer documentation. “We will generate in any language they want,” Arora said. “We are essentially trying to serve all the stakeholders, from the no-code people to the data engineers, on the same platform.”

Image: Pixabay