Labels

What are labels?

Labels let you categorize threads so your team can route, filter, and prioritize work effectively. For example, you might use labels like bug, billing, or feature-request to keep things organized.

Labels can be nested — you can create a parent label (e.g. billing) with child labels underneath it (e.g. billing > refund, billing > invoice).

Applying labels to a thread

To add a label to a thread, open the thread and click the label picker. Start typing to search for a label, then click it to apply it. You can apply multiple labels to a single thread.

Creating a label from the thread view

You can create a new label without leaving the thread you're working on — no need to go to Settings first.

1. Open the label picker in the thread view.
2. Type the name of the label you want to create.
3. A Create "[name]" option will appear at the top of the list — click it.
4. An inline form will appear with the label name pre-filled. Edit it if needed.
5. Optionally, select a parent label to nest the new label under an existing one.
6. Click Create label (or press Enter) to create it. The label will be created and automatically applied to the thread.

Press Escape at any point to cancel without closing the label picker.

Note: Labels created this way have AI auto-labeling turned off by default. You can change this in Settings > Labels.

Managing labels

To rename, delete, or configure labels (including enabling AI auto-labeling), go to Settings > Labels.

Best practices

Structuring your labels

• Keep the top level broad, use nesting for specificity. A flat list of dozens of labels becomes hard to navigate. Instead, use parent labels for categories (e.g. billing) and child labels for the specific cases within them (e.g. billing > refund, billing > failed payment).
• Name labels for what happened, not how you feel about it. Labels like urgent or annoying don't age well and are hard to act on. Labels like bug, data-loss, or enterprise-blocker are specific and searchable.
• Avoid overlap. If two labels frequently get applied together, consider merging them or making one a child of the other. Overlapping labels dilute the value of your reporting and make AI labeling less accurate.
• Audit regularly. Labels created on the fly can accumulate quickly. Review your label list in Settings > Labels periodically and consolidate or remove labels that are no longer useful.

Writing descriptions that help AI label accurately

When AI auto-labeling is enabled for a label, Plain uses the label's name and description to decide whether to apply it to a thread. A well-written description is the single most effective way to improve labeling accuracy.

• Be explicit about what qualifies. Instead of a vague description like "billing issues", write something like: "Apply when a customer is asking about a charge, invoice, refund, failed payment, or subscription change." The more concrete the criteria, the more consistently the AI will apply it.
• Include examples of what the thread might say. Mention the kinds of phrases or topics that are a signal. For example: "Threads where the customer mentions they were charged incorrectly, wants a receipt, or is asking why their card was declined."
• Describe what does NOT qualify. If a label is frequently confused with another, add a negative example. For instance: "Do not apply if the customer is just asking about pricing — use the 'pricing' label instead."
• Keep it to 2–4 sentences. Descriptions that are too long or too vague are harder for the AI to act on. Aim for precise and scannable.