October 2026 Summaries

No summary generated yet.

Sigma Assistant build mode is now generally available, enabling users to create and modify Sigma dashboards and simple warehouse-connected applications through natural-language requests. It can generate workbook elements such as charts, filters, layouts, pivot tables, conditional formatting, input tables, forms, and automated actions using live warehouse data while honoring existing data models, metrics, lineage, and permissions. Users can build dashboards from a single prompt, review proposed structures before changes are made through plan mode, create writeback workflows such as forecasting or approval apps, and extend existing workbooks rather than creating separate copies. Sigma positions the feature as a way to reduce dashboard development time, make advanced analytical and app-building capabilities accessible without prior product expertise, and preserve governance through editable native workbook outputs. Planned enhancements include using shared assets to match existing workbook styles and allowing organizations to create reusable custom skills, while access requires a supported AI provider and relevant Assistant and workbook permissions.

Sigma’s MCP server and CLI let MCP-compatible AI assistants such as Claude, ChatGPT, Cursor, and Codex interact with Sigma using existing user permissions, enabling natural-language analysis, application building, and administrative tasks without custom integrations. MCP provides access to governed data, search, querying, and workbook-building capabilities, while the CLI exposes Sigma’s REST API for actions including user and team management, credential audits, and activity reviews, with confirmation required for destructive operations. In demonstrations, assistants located relevant data sources, built editable Sigma applications and dashboards in minutes, and used MCP to verify that generated dashboards matched source-data analyses. The workflow relies on existing metadata and benefits from custom skills that provide business context and design standards, while code-driven changes have a different version-history model than UI edits; production-tagged workbooks remain insulated from such changes, and Sigma plans more Git-like history features in the future.

Sigma’s September 2026 releases expand how its AI agents, dashboards, and reporting tools operate across business workflows. Agents built in Sigma can now run on schedules, through the REST API, or via Claude and ChatGPT integrations while retaining their instructions, connected tools, and user-specific data permissions; new management capabilities provide organization-wide visibility into agent usage, execution logs, costs, evaluations, and behavioral drift. Sigma is also integrated with OpenAI’s Data agent and Codex, allowing ChatGPT users to create, refine, and publish live-data Sigma dashboards from prompts under existing access controls. An upgraded Sigma Assistant asks clarifying questions for ambiguous requests, exposes its reasoning and generated SQL, retains conversation history, and can access all sources a user is authorized to query. Reporting additions include a public-beta PowerPoint export that creates permission-aware, warehouse-refreshed slide decks and a workbook-to-report converter that produces paginated reports without altering the original workbook.

Agentic automation in data analysis uses AI agents to pursue human-defined goals across detection, analysis, decision-making, and optionally action, extending beyond traditional dashboards, fixed scripts, RPA, and chatbots that stop at reports or answers. It can help analytics teams accelerate decisions, consistently investigate both aggregate trends and record-level drivers, and address recurring low-priority exceptions that human reviewers often overlook. Effective deployment depends on governed data access that inherits user permissions, least-privilege limits on permitted actions, and detailed audit trails capturing triggers, data, tools, context, decisions, and human intervention. The article recommends beginning with a narrowly scoped decision, retaining human approval for high-impact actions, monitoring outcomes, and expanding autonomy only after demonstrating accuracy and auditability. It presents Sigma as a platform for governed agentic workflows, with agents configured in workbooks around defined data scopes and approved actions, warehouse-based security, scheduled or human-reviewed execution, and writeback auditing through Input Tables.