ThoughtSpot agents aimed at automating the analytics lifecycle – News

ThoughtSpot Inc. today introduced a suite of business intelligence agents designed to automate major components of the analytics workflow, spanning data modeling, dashboard creation, embedded development and real-time analytical queries.

The four agents — SpotterModel, SpotterViz, SpotterCode and an upgraded Spotter 3 — extend the company’s push into what it calls “agentic analytics,” a model that uses semi-autonomous software agents to reduce manual work for data engineers and business users.

“Our customers spend a lot of time building pipelines, semantic models and dashboards but not really providing insight,” said Francois Lopitaux, senior vice president of product management at ThoughtSpot.

The new agents are designed to pair with different personas involved in analytics operations, such as analysts and developers. “We want to provide a friend in the business that helps them to perform their tasks better,” Lopitaux said.

Embedded development

ThoughtSpot said SpotterModel simplifies semantic modeling by proposing tables, joins and logic based on natural-language instructions. Though agents are capable of full autonomy, their primary role is as advisers, Lopitaux said. “There is a strong component of human in the loop,” he said. “SpotterModel comes back with a suggestion but the human needs to validate it.”

SpotterViz automates dashboard creation, including layout, design and the selection of key performance indicators. While ThoughtSpot executives have been vocal in advocating for users to move beyond dashboards, “people are still using them and we want to make it easy for them to transition to the next generation,” Lopitaux said.

For developers, SpotterCode provides code generation and embedded analytics support within popular integrated development environments. The agent understands each developer’s project context and can generate code ThoughtSpot code derived from the company’s software development kit. “It’s like having an expert sitting next to you,” Lopitaux said. “When you need it, you can tap into its knowledge or even ask it to write the code.”

At the center of everything is Spotter 3, the newest version of ThoughtSpot’s agentic analyst. Spotter allows users to ask analytical questions and get answers without building dashboards or using structured query languages.

Lopitaux said the new version system can answer complex queries, integrate with third-party software using the Model Context Protocol and generate Python code when needed. “You can plug Spotter 3 directly into a company’s SaaS applications,” he said. “Spotter is able to not only use the data that is available to ThoughtSpot, but also reach out to bring additional information.”

To guard against errors and hallucinations, Spotter 3 applies iterative reasoning to evaluate and refine its own results. “Every time you ask a question, it explains its chain of thought,” Lopitaux said. ThoughtSpot also uses its own search-token technology to generate SQL rather than relying on large language models.

The agents don’t yet coordinate autonomously with each other, though they operate on a shared metadata layer. Lopitaux said cross-agent cooperation is on the company’s roadmap. The company is also developing customizable agents that can perform specialized tasks, such as identifying customers likely to churn.

Spotter 3 and the new agents are scheduled for release early next year.

Image: News/Dreamina