Skip to content

How does Krina classify @mentions?

Every @mention is classified automatically across three independent dimensions as soon as it’s posted. There’s nothing to configure — the AI handles it.

Meaning
RESPOND — A direct question or request needs a reply
REVIEW — Feedback or review has been requested
APPROVE — Your sign-off is needed before work can proceed
FIX — Something is broken and needs correcting
BUILD — A new feature or enhancement needs building
WATCH — An ongoing situation to keep an eye on — no action needed yet
NOTE — A completed event or notification — informational only

Note: @mentions of yourself are not processed by Krina. To track personal items, use Personal Reminders (Track This) instead.

ValueTimeframe
CRITICALRespond within < 1 hour — active incident or critical blocker
HIGHRespond within < 4 hours — time-sensitive with near-term deadline
NORMALRespond within < 3 days — standard work with reasonable deadline
LOWNo deadline — can wait indefinitely

Urgency badges include a short hint describing what’s at stake, so you don’t have to open the item to know why it’s marked the way it is.

ValueMeaning
CUSTOMERAn external party — users, clients, patients, students
REVENUEA financial outcome — sales, billing, contracts
TEAMAn internal group or department
INDIVIDUALOne person only

Each classification includes a short, AI-generated explanation. Hover over the Action, Urgency, or Impact badge on any item to see the specific reasoning behind that part of the classification.

The short “hint” shown next to the urgency badge is also AI-generated, and is written to describe what’s at stake — it never includes timing words like “today” or “ASAP,” since the timeframe is already shown separately. The hint reflects the item as it was originally classified and doesn’t change even if urgency is later escalated (see Understanding Urgency & Escalation).

Group mentions — how Krina handles multiple people

Section titled “Group mentions — how Krina handles multiple people”

When you mention multiple people in one comment, Krina classifies the mention the same way for each person. Each mentioned person receives their own action item in their inbox with the identical Action, Urgency, and Impact classification based on the comment content.

@Alice @Bob — Can you review the API changes and approve the deployment?
We need this live by Friday.

This creates separate action items for each person with the same classification:

  • Action: APPROVE (formal sign-off is requested)
  • Urgency: HIGH (deadline this week)
  • Impact: TEAM (affects the whole team’s release)

Both Alice and Bob see this classification in their respective inboxes. They can work together on it, coordinate their responses, or one can handle it independently. Resolving it in your inbox marks it complete for you; it doesn’t automatically resolve it for the other person.

Use caseDo thisWhy
Ask specific people to collaborate@Alice @Bob can you both review?Creates one mention, both see it
Ask different people different thingsPost two separate comments with individual mentionsCreates two separate mentions with different actions
Ask a team@team-nameMentions the entire group at once

If you mention 5 people but write “please take a look” without any urgency signal, all 5 see it classified as NORMAL/REVIEW. The mention isn’t repeated for each person — it’s one item that multiple people have responsibility for. This makes it clear who’s been looped in and avoids cluttering inboxes with duplicates.

Krina processes @mentions across eight surfaces:

  • Jira issue comments
  • Jira issue descriptions
  • Confluence page bodies
  • Confluence blog posts
  • Confluence inline comments
  • Confluence footer comments
  • Jira Service Management ticket descriptions
  • Jira Service Management ticket comments

Note: @mentions inside Confluence databases or whiteboards aren’t currently picked up — Confluence’s API doesn’t expose that content the same way it does for pages and comments.

Note: Editing a comment after it’s posted triggers reclassification automatically, in both Jira and Confluence — but only if the content actually changed. Cosmetic edits (formatting, whitespace) don’t trigger a new classification; the original stands.

If classification doesn’t complete (for example, if the AI doesn’t respond in time), the item is marked as failed rather than left unclassified. You can trigger a retry by editing the comment again.

Privacy PolicySivect