ThoughtSpot is giving administrators the full Spotter trace—and a benchmark to test it

The agent gets more context—and administrators get the trace

ThoughtSpot’s planned 26.11 Cloud release turns Spotter into a more observable analytics agent. The company says Spotter on Liveboards will become generally available, grounding conversations in a Liveboard’s tabs, visualizations, effective filters and time ranges. Multi-step work such as forecasting, “why” analysis and charting will remain in one conversation, with each step tracked across Spotter and the Liveboard. The release is scheduled for October 28, according to ThoughtSpot’s current release page. (ThoughtSpot)

The consequential change is on the administrative side. ThoughtSpot says administrators will be able to open a usage dashboard or search all Spotter conversations, preview a conversation, download the list as CSV, and inspect full message histories, tool calls and execution timings. Filters include user, agent, data source and date range. The same release is also set to make an in-product benchmark platform generally available: authorized users can define a sample question, specify a correct answer in text or SQL, run it and compare selected runs for latency and accuracy. (ThoughtSpot)

This closes two production gaps

Those controls address two different operating problems. The benchmark gives teams a repeatable acceptance test before changing models, instructions or data sources. The conversation trace then helps explain a failure after deployment: operators can inspect which source and tools the agent used, rather than judging only the final prose. That combination is more useful than a generic thumbs-up metric because it connects expected answers to execution evidence.

ThoughtSpot is also widening what Spotter can do. Its current 26.10 release made Spotter Analysts and their chat-based builder generally available, while adding a connector marketplace with more than 100 third-party integrations. The same release added connector sharing by user and group, instead of exposing a configured connector to every user across all organizations. (ThoughtSpot 26.10 release notes)

The governance question moves to the administrator

Greater observability creates a second control surface. Conversation histories can contain business questions, returned values and tool activity. ThoughtSpot’s preview page describes who can open monitoring and what they can inspect, but it does not state retention periods, masking behavior or whether conversation access can be split from broader Spotter-management privileges. (ThoughtSpot)

Before enabling monitoring, operators should therefore decide who may review chats, how exported CSV files are handled and which benchmark questions may contain production data. They should also test the effective Liveboard filters and permissions used by benchmark identities. A correct answer under an administrator’s access can still be the wrong acceptance test for a restricted business user.

ThoughtSpot’s direction is clear: production analytics agents need both evaluation before release and traces after release. The next task for customers is to govern the observers as carefully as the agent.