Standard, Custom, or Free Language: Three Ways to Classify What Matters

Not all sensitive data looks the same, and that is why single-mode classification fails. A tool that only recognizes credit-card numbers will miss the merger memo. A tool that only matches your keyword list will miss everything you didn't think to list. Real environments need more than one lens.

Capsule Insight classifies in three layers, and they work together.

1. Standard labels — the known categories

Out of the box, Insight identifies the sensitive data every regulated organization must account for: personally identifiable information (PII), financial data, source code and credentials, medical records, legal documents, and intellectual property. These are the categories auditors ask about, detected across every connected system without setup.

2. Custom labels — your specific terms

Every organization has data that only matters to it: internal project codenames, a specific client identifier format, a product line under NDA. With custom labels, you define the keywords that matter to your business, and Insight finds them everywhere they appear — no dependence on a generic ruleset.

3. Free-language labels — describe it, and it's found

This is where classification stops being rigid. Through an AI Agent, you can define what you're looking for in plain language — "contracts that mention early termination," "documents referencing an unannounced acquisition." No regular expressions, no manual rules. You describe the intent, and Capsule detects it at scale.

Why three layers, not one

Standard labels give you coverage of the obvious. Custom labels capture what's specific to you. Free language reaches the data that has no fixed pattern at all — the memos, the notes, the context that rigid tools never catch. Together they turn "we scan for a few data types" into "we understand what our documents actually contain."

And because every classified file carries a risk score, the output isn't just a longer list — it's a clearer one.

Classification is only useful if it matches how your data really looks. Three layers make sure it does.

Want to see how Insight classifies your own documents? Request a demo →.