ThoughtSpot domain-specific Spotter agents target AI success

Context is key

While the reasons many AI initiatives fail to make it into production vary, a prominent one is that tools such as GenAI chatbots and agents aren't being fed relevant data to carry out their purpose.

ThoughtSpot's industry-specific agents could effectively address the high failure rate of AI projects by bridging the context gap that limits general AI.

For example, a customer service agent needs to access all data appropriate to an individual customer and any issues that customer is facing, and the access has to be immediate. If the agent can't discover the right data in real time, it won't be of any use when a customer calls with a complaint.

Similarly, agents designed to enable users to reach insights regarding travel bookings or optimizing supply chains need to access relevant data to have the contextual awareness to deliver appropriate outputs.

Recently, Databricks introduced Instructed Retriever to give agents more contextually relevant data by improving data retrieval accuracy. MongoDB similarly addressed improving data retrieval by launching new embedding and reranking models. And numerous vendors such as GoodData and Pentaho have added or improved semantic modeling capabilities to make data more consistent and discoverable so AI tools can call on relevant data.

ThoughtSpot, which has provided a semantic layer since its inception and introduced Spotter Semantics on March 12 to provide an AI-powered context layer between data sources and agents, is addressing the need for contextually appropriate data by adding agents with prebuilt domain-specific expertise.

Spotter for Industries agents, which build on the capabilities of Spotter Semantics, include built-in understanding of domain-specific logic, regulations related to the industry and key performance indicators relative to each industry. In addition, they include capabilities that streamline some of the work unique to each industry.

For example, Spotter for Healthcare and Life Sciences automatically connects unstructured data such as research notes in documents with financial, clinical and claims data sources. Meanwhile, Spotter for Media and Telecommunications autonomously links streaming logs with audience sentiment data.

"Spotter for Industries … is like using a small language model instead of an LLM," McKnight said. "It lets you mash together messy, unstructured data with your official numbers to give a full picture. Better yet, it proactively spots real-world problems before they blow up. These are all significant."

Matt Aslett, an analyst at ISG Software Research, noted that many data management and analytics vendors have historically provided industry-specific capabilities to make it faster and easier for enterprises in given industries to combine applications with the relevant data for their business needs. For example, Databricks and Snowflake each have numerous domain-specific versions of their data platform while SAS and SAP are among BI vendors that offer tailored capabilities.

Now, as data management and analytics providers expand to support GenAI and agentic AI, it's a natural evolution to extend industry-specific capabilities to AI tools such as ThoughtSpot's Spotter for Industries agents, according to Aslett.

"By providing agents with this industry-specific information, ThoughtSpot is potentially reducing the level of training, prompting and tailoring required to ensure generic agents are provided with industry-specific context, as well as trusted enterprise data," he said.

In addition, although ThoughtSpot is not the only vendor offering agentic AI capabilities tailored to the needs of enterprises in specific industries, Spotter for Industries is distinguished by the depth of their contextual awareness and the number of different industries ThoughtSpot is addressing, Aslett continued.

"While ThoughtSpot is not alone in delivery of industry-specific agents, the breadth of industries being addressed is a differentiator, along with its agentic reasoning layer and semantic modeling language capabilities," he said.

Although a valuable addition for ThoughtSpot customers and a potential differentiator for ThoughtSpot as vendors continue to add capabilities aimed at enabling agentic AI development, seeing that generic AI models lack enough context to enable businesses to reach useful insights provided ThoughtSpot with the impetus for developing Spotter for Industries, according to Francois Lopitaux, the vendor's senior vice president of product management.

"While generic AI models are excellent at broad language tasks, they often lack the literacy in specific regulations, workflows and terminology critical to specialized industries," he said. "We developed Spotter for Industries to bridge this 'context gap' and ensure all critical industry-specific data is wrapped into the semantic layer."

Next steps

As ThoughtSpot plans future product development, adding more agentic capabilities that automate its platform is an ongoing emphasis, according to Lopitaux.

"Our focus remains on building for the autonomous future, where agents independently detect, decide and execute workflows, acting as a true extension of your team," he said. "We see this as a world in which everyone gets an agent, workflows are autonomous, and every dashboard becomes an AI native data app."

McKnight, meanwhile, suggested that ThoughtSpot focus on upgrading the capabilities of its Spotter for Industries agents given that enterprise demands continue to grow and the new tools will need to meet those demands to enable customers to operationalize their AI projects.

"ThoughtSpot will likely have its hands full ruggedizing these industry templates and its multi-modal architecture," he said. "Successfully operationalizing this shift will eventually allow them to move from insight to action -- the next frontier in AI."

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