To auto-apply a retention label to documents based on a specific file naming pattern (like abc_XX123.docx), you need to use a detection method that can recognize patterns.
Retention label: This is the outcome (what you apply), not the detection mechanism. It cannot itself identify files.
Trainable classifier: Used for identifying content based on text meaning/semantics (e.g., HR documents, contracts). It is not suitable for structured patterns like file names.
Sensitive info type (SIT): Best option. Custom SITs can be created with regular expressions to match naming formats (such as xxx_XX999). Once the SIT is defined, you can configure an auto-apply retention label policy to apply the "Project Documents" label when the SIT is detected.
Reference: Auto-apply a retention label