datascalehr’s AI works with five categories of data.

Personal data (worker, HR and payroll data) is processed by internal algorithms only and is never sent to an LLM or stored in KMod™.

Header data (column and table names), contextual data (country, grouping, pay run dates) and categorization data (data type, format, validation rules) feed the mapping suggestions and may be sent to an LLM.

Behavioral data, the expert’s confirmation or rejection of each suggestion, is the training signal that KMod™ stores.