What is data?
Data is the recorded facts, measurements, and events that a business stores so that people and software can use them later. Every system produces it: the ERP records orders and inventory movements, the CRM records calls and quotes, machines emit readings, and email and document systems hold contracts and correspondence.
Data comes in three broad shapes. Structured data sits in rows and columns with fixed types, such as a general ledger table. Semi-structured data carries its own labels but no fixed layout, such as a JSON message from an API. Unstructured data has no predefined fields at all, such as PDFs, images, voice recordings, and free-text notes. Most companies hold far more unstructured data than structured, and most of it has never been analyzed.
Why data matters
An AI model can only learn patterns that exist in the data it is trained on, and it can only answer questions about data it can reach. A demand forecast for a regional distributor is only as good as the order history behind it: missing months, inconsistent product codes, and returns booked as sales all show up in the forecast. A document assistant will quote an outdated policy if the outdated version is what it was given.
AI programs therefore spend most of their effort on data work: finding it, cleaning it, defining what each field means, and controlling who may see it. The terms data quality, data governance, and data lineage describe that work, and a dataset is the packaged slice of data that a specific model is trained or evaluated on.