November 2026 Summaries

Management reporting software creates recurring internal reports that help managers and executives evaluate performance using KPIs, variance and trend analysis, explanatory narrative, and drill-down detail, distinguishing it from externally focused statutory reporting governed by GAAP or IFRS. Common applications include budget-to-actual tracking, board and regulatory reporting, departmental KPI monitoring, and forecast and pipeline reviews, all of which benefit from timely, traceable data drawn from consistent sources. The article argues that traditional extract-based spreadsheets and static reports can introduce delays, errors, governance risks, and difficulty answering follow-up questions, while live warehouse-connected reporting can improve currency, auditability, standardization, and analyst productivity. It also describes how AI interfaces and agents can support natural-language analysis, automated variance commentary, and scheduled report preparation when they operate on governed data with inherited permissions. Sigma is presented as a platform offering pixel-perfect exports, live warehouse queries, AI-assisted analysis, automated agents, and warehouse-based row-level security for management reporting workflows.

AI is reshaping business intelligence by moving beyond static dashboards toward natural-language querying, automated insight and anomaly detection, predictive forecasting, and agentic workflows that can take governed actions. Legacy BI is limited by predefined questions, analyst-dependent follow-up, and disconnected workflows, while AI-enabled systems can answer conversational questions, identify emerging patterns, model scenarios, and connect findings to tasks, approvals, or planning updates. Successful implementation depends less on models alone than on live, governed data access, a shared semantic layer with consistent metric definitions, inherited user permissions, auditability, and organizational change management. The recommended approach is to connect AI directly to governed data at query time, begin with a narrow measurable use case, pilot with the intended business team, monitor accuracy and metric consistency, and scale in controlled phases. Sigma presents its platform as a warehouse-native layer that supports these requirements by querying live data in systems such as Snowflake, BigQuery, Databricks, and Redshift, while providing governed AI assistants and configurable agents that can analyze data and initiate approved downstream actions.

Sigma has released a public beta connector for MotherDuck that lets all Sigma customers query MotherDuck databases live to create workbooks, data models, dashboards, reports, and AI-driven analyses without data extraction or pipelines. The integration exposes MotherDuck tables, views, and AI functions such as PROMPT and PROMPT_JEV, while Sigma provides spreadsheet-style exploration, custom SQL, governed data models, natural-language assistance, agents, sharing, embedding, and permission controls. Administrators can configure the connection through MotherDuck’s Postgres endpoint using a Read Scaling access token, then grant users access to begin building. The beta is read-only, meaning writeback and features including input tables, CSV uploads, Python code, materialization, cloud-storage exports, SSH tunnels, OAuth, and some role-switching capabilities are not yet supported, though writeback is planned.

Sigma has introduced headless access for its governed AI agents, allowing them to operate outside Sigma workbooks through MCP-compatible assistants such as Claude, ChatGPT, Cursor, Codex, and CoCo, or through custom applications using the Sigma REST API. Agents retain the instructions, data sources, and user-level permissions configured in Sigma, with each request executed under the calling user’s existing access rules, including row-level security. The MCP server uses OAuth to let assistants discover and query available workbook agents, while the REST API supports embedded interfaces, automations, streaming responses, full execution traces, and structured JSON outputs for uses such as charts and downstream workflows. Demonstrations showed users asking sales questions through Claude and building an embedded agent hub that lists accessible agents, visualizes responses, and can connect to third-party MCP services such as Notion to perform actions. Sigma emphasizes that agents are managed centrally in published workbooks, enabling organizations to reuse the same governed setup across interfaces without rebuilding or maintaining separate security models.

Sigma Tables are now in public beta, offering editable, warehouse-native tables that can be shared across Sigma workbooks, applications, and external systems while keeping the data warehouse as the system of record. Available initially for Snowflake, with Databricks and PostgreSQL support planned, they enable teams to create shared workflows in which changes made in one app immediately appear in others. Public beta additions include input columns for adding editable comments or values beside read-only source data, content validation to protect dependent workbooks during schema changes, append-only audit logs of edits, and actions that let apps insert, update, or delete rows through buttons or automated triggers. Sigma positions the feature for collaborative finance budgeting, retail inventory and purchase-order adjustments, and marketing budget management, while private beta users have applied it to recursive inventory planning, shared analytics data models, and hierarchical review feedback. Planned enhancements include row- and column-level security, feature parity with Input Tables, and support for additional warehouses such as BigQuery, Redshift, and ClickHouse.

Sigma describes using Snowflake interactive warehouses with existing Sigma connections to deliver low-latency, high-concurrency access to live governed data for dashboards, embedded analytics, and application read paths without creating extracts or duplicating data. Administrators can route an entire connection, individual workbook, or user groups to an interactive warehouse, making the approach suited to frequent, selective queries such as filtering and drilldowns while retaining standard warehouses for exports, broad scans, Python workloads, write-back, and other longer-running tasks. Interactive warehouses impose a five-second query limit, after which Snowflake automatically reruns queries on a configured fallback warehouse, with OAuth users requiring access to both. Snowflake’s existing role-based permissions, row access rules, masking policies, and auditing continue to apply because data remains in Snowflake. The recommended adoption path is a one-workbook pilot measured against realistic usage and concurrency, while operational guidance includes maintaining a warm cache, attaching frequently used tables for proactive caching, monitoring remote reads, and sizing warehouses for both data working sets and peak demand.

AI agents for supply chain and demand planning are governed systems that continuously read live ERP, warehouse, point-of-sale, and transportation data to detect changes such as demand spikes, delivery delays, or inventory risks, then make or recommend authorized adjustments to forecasts, replenishment orders, and exception workflows. Unlike spreadsheets, dashboards, and conventional recommendation engines, they can reduce latency and manual re-entry by writing scoped, auditable actions back to systems of record while reserving higher-impact decisions such as network redesigns and contract negotiations for human oversight. Their highest-value applications include continuous demand sensing, automated replenishment, and pre-triaged supplier or logistics exceptions, where frequent and largely reversible decisions otherwise consume substantial planner time. Reliable deployment depends on accurate, current data, predefined execution limits and approval thresholds, and detailed logs explaining each action and its underlying data. Sigma positions its warehouse-native platform as a runtime for such agents, using existing warehouse permissions, row-level security, lineage, live queries, and approval-enabled writeback capabilities to support governed planning workflows.