ThoughtSpot Embedded update boosts agentic AI capabilities

Infusing AI

Before generative AI enabled true NLP, BI was limited to trained experts and self-service users with data literacy skills. As a result, for decades, only about one-quarter of employees within organizations used data to inform their work.

Embedded analytics was one way some organizations successfully enabled more workers to benefit from BI, providing them with data-informed insights within their normal workflows rather than forcing them to learn a complex analytics platform.

Now, embedded analytics is expanding to include AI capabilities, such as freeform natural language queries, directly within applications like Salesforce and Workday, as well as tools custom-built by an employee's organization.

Like Farmer, David Menninger, an analyst at ISG Software Research, called the ThoughtSpot Embedded update "an evolution" because it advances the AI capabilities users can embed.

"As software providers create additional agentic capabilities, it's important to make those capabilities available in the embedded version as well," he said. "Other new features … round out the edges, making it easier to brand the embedded capabilities consistently with the application in which it's being embedded."

ThoughtSpot Embedded is now built on four core pillars, according to Francois Lopitaux, the vendor's senior vice president of product management. They include the following:

• Embedded intelligence refers to the infusion of AI and BI in applications through ThoughtSpot's Visual Embed SDK and REST API.
• Native experiences, including an Integrated Developer Playground and Mobile SDK that enable developers to create web and mobile applications that use ThoughtSpot's capabilities but are designed with the look and feel of their organization's tools.
• Seamless workflows encompass ThoughtSpot's model context protocol server, allowing users to extend Spotter's source data to AI platforms such as ChatGPT and Anthropic's Claude.
• An open ecosystem that includes ThoughtSpot's cloud-agnostic architecture layered on top of the vendor's agentic semantic layer and built-in governance capabilities to help ensure that data can be trusted.

"The four pillars make sense," Menninger said. "They address real issues that have been obstacles to reaching more of the workforce with analytics. They've made good progress on embedding and native experiences."

Meanwhile, determining what features to include in ThoughtSpot Embedded was driven by customer feedback, according to Lopitaux.

"We already had a lot of customers using our solution, but all these customers wanted more," Lopitaux said. "Mobile SDK was one of the biggest requests that they wanted to add. Also, to be able to embed Spotter in their own applications is definitely top of the pile because they want to provide agentic experiences to their customers."

One such customer is Navan, a corporate travel and expense management company. Using ThoughtSpot Embedded, Navan added ThoughtSpot's AI capabilities to its website so that its customers can now query data using natural language. For example, Navan users can ask which of their employees traveled the most during the previous 12 months, where and why they traveled, and how much those employees spent on travel to gain insights into how the travel budget is used.

"[Embedding AI] helps to avoid the proliferation of dashboards because you never know what the follow-up questions will be," Lopitaux said. "When you have a dashboard, you need to send it to an analyst or [IT] to update a visualization. We are breaking that by allowing people to directly ask a question and get an answer right away."

Spotlight on Spotter

While the ThoughtSpot Embedded Update is based on four pillars, the latest version of Spotter enables the agent to act as an MCP host so it can connect to unstructured data sources such as Slack, improve the agent's integrations with large language models to increase its intelligence and adds Research Mode to better reason and validate outputs by breaking complex requests into multi-step questions.

"The MCP integration stands out," Farmer said. "I also rather like Research Mode, which lays out a multi-step reasoning process for analytics."

However, Spotter 3 represents ThoughtSpot's first foray into offering agentic reasoning capabilities, and there remains room for improvement, he continued.

"This is in its early stages," Farmer said. "I want to see this multi-step process become more interactive with users confirming the research plan or nudging it in a different direction after review."

Menninger similarly highlighted the value of Spotter 3 acting as an MCP host and the agent's Research Mode.

"The MCP Server Connection makes it easy for ThoughtSpot users to incorporate their analytics into other generative and agentic AI tools," he said. "The research capabilities will support deeper analyses and reports based on ThoughtSpot analytics."

Together, while improvement such as adding an interactive component to Spotter's multi-step reasoning capabilities is needed, the ThoughtSpot Embedded and Spotter 3 advancements serve the needs of ThoughtSpot's users, according to Farmer. In addition, he noted that the updates could make the vendor's embedded analytics capabilities more competitive with those of embedded BI specialists.

"In all, this is an excellent release which addresses many of the weaker points in ThoughtSpot's embedded story and will make them much more competitive," Farmer said. "Other vendors -- I am thinking of Sisense in particular -- will have to take note."

Menninger, meanwhile, noted that while valuable for ThoughtSpot users and keeping the vendor competitive, the update does not necessarily help ThoughtSpot stand apart from vendors such as Tableau that also provide embedded analytics and agentic AI capabilities.

"It is really hard for software providers to differentiate their products in a mature market like business intelligence," he said. "Nearly all the vendors are racing toward the same goals of making it easier to use analytics and making it easier to integrate analytics into business processes. … Products like Tableau Next and others are similar to ThoughtSpot Embedded and Spotter."

Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than 25 years of experience. He covers analytics and data management.