Ask a question, get a grounded answer in under two minutes

The AI Copilot can answer questions about any data available in the Demand OS platform: current forecast values and recent changes, inventory positions and days of cover by SKU and location, open purchase orders and expected receipts, demand sensing alerts and the signals that triggered them, scenario assumptions and inventory impacts, and factor attribution from the Explainable AI module. It cannot answer questions about data outside the Demand OS data layer — external market data or ERP data not yet ingested.

The AI Copilot uses a retrieval-augmented approach — it queries the live planning data to construct its answer rather than generating a response from model internals. Every answer includes the specific data points it drew from, and planners can drill into the source record from the Copilot response. The system is designed to say "I don't have data to answer that" rather than produce a fluent-sounding but unverifiable response.

The Copilot improves its query routing and answer formatting based on usage patterns within a customer environment, but it does not use planner interactions to update its underlying model in ways that could change the factual basis of answers. The factual grounding always comes from the live planning data, not from accumulated conversation history.

In the current implementation, the AI Copilot is an information and analysis tool — it surfaces data, highlights exceptions, and explains patterns. It does not take autonomous actions such as approving forecasts, generating purchase orders, or modifying planning parameters. All planning actions require planner review and confirmation in the planning workbench.

The AI Copilot is available to all licensed Demand OS users. Access to specific data sets within the Copilot follows the same permission structure as the planning workbench — a user who does not have access to a product category in the workbench cannot query that category through the Copilot.