Anomaly AI

Every Monday the weekly spreadsheet takes a little longer to open, and someone asks where the figure in cell F42 came from. Reports built in spreadsheets slow down as the data grows, and it gets hard to see how a figure was worked out. In Anomaly AI you upload or connect the data and ask a question in plain language. The AI then writes a small program in SQL or Python, two common languages for data work, and that program does the calculation.

You can open the code and the rows behind every figure and run it again. Before the analysis starts, Anomaly checks the data for duplicates and missing values. It also flags values far out of line with the rest. Metric definitions and business rules you set once stay in place for the next reporting cycle.

Data comes from uploads or from connected sources, for example GA4 and BigQuery. The security page says customer data is never used to train AI models. Through an MCP server, AI assistants like Claude or ChatGPT can send questions to Anomaly and receive the answer with links to the files.

Results become dashboards that refresh when new data arrives, or files in Excel, PowerPoint, Word and PDF. Reports can also go out by email on a fixed schedule.