I replaced Excel's Copilot with a local LLM, and spreadsheet work finally feels private

The described local AI assistant for Excel leverages Qwen3.5-9B, a specific LLM, managed by Ollama for local execution and connectivity. An open-source Excel add-in then provides the interface, enabling the model to access spreadsheet data and perform actions directly within the application. This configuration addresses concerns about data privacy by ensuring that workbooks remain on the user's computer, avoiding the need to transmit sensitive information to cloud-based AI services.

This local setup offers practical assistance for common Excel tasks, including generating formulas from natural language descriptions, cleaning text data, and applying basic formatting. For example, the agent can translate a request like "calculate revenue from quantities and prices" into an Excel formula, or help categorize customer comments. These capabilities are noted to provide considerable overlap with the everyday functions of cloud-based alternatives for many users.

While this local approach prioritizes data privacy and provides core utility, it also presents trade-offs. The report indicates that cloud-based services like Copilot offer a broader feature set, including advanced functions such as PivotTable creation and access to external information, which the local model does not replicate. Users considering a local solution would weigh the appeal of keeping their data on-device against more expansive capabilities of cloud-integrated AI tools.