The Action Inbox Playbook

Most HR ops leads live across seven browser tabs: a compliance dashboard, a debt tracker, a policy library, an onboarding checklist, an audit log, an email client, and a shared spreadsheet for the things that fall through the cracks. The Action Inbox at /hrdashboard/inbox exists to collapse those seven tabs into one queue.

Every workflow in Clarus HR feeds the inbox. The compliance scan emits a card for every drift item it finds. The debt scan emits a card for every policy contradiction. Policy rollouts emit cards when an acknowledgment is late or an exception is unresolved. New hires emit cards when an onboarding step is overdue. The inbox is the single workbench: open it in the morning, close items through the day, and aim for inbox zero by 5pm.

Every card carries one of four severity labels, color-coded so you can triage at a glance:

• CRITICAL (rose): Imminent legal, financial, or safety exposure. Examples: an I-9 unverified past day three, a breach notification window expiring, a manager flagged for a complaint pattern.
• WARNING (amber): Important but not yet harmful. Examples: SOC 2 evidence captured but expiring within 30 days, an acknowledgment overdue by under a week.
• NOTICE (blue): Routine work that needs attention but has no deadline. Examples: a policy version published with new audience filters, a manager briefing ready to review.
• INFO (gray): Reference items that document changes made elsewhere. Examples: a new jurisdiction added to the engine, a global requirement updated.

Sort by severity descending and the queue triages itself. Most teams work CRITICAL and WARNING in the morning, NOTICE through the afternoon, and leave INFO items for a Friday review.

Each card is also assigned one of four type buckets that say what kind of work the card is asking for:

• urgent_escalation: Someone needs to act right now and escalate to a manager or counsel. The card carries a dot indicator and a deadline.
• needs_action: Routine work the inbox owner can finish without a handoff. Most compliance and debt items land here.
• needs_approval: A pending decision waits on a named approver. The card lists the approver role.
• informational: No action required. Read, dismiss, move on.

Two fields on every card make routing fast: recommended_action is an AI-drafted next step you can run as-is, and affected_employee_ids is the list of employees in scope so a manager can review impact before approving.

Closing a card uses POST /v1/hr/inbox/{id}/close. The modal asks for a reason string before it confirms. The reason is not optional UI polish; it is the audit log entry that survives a discovery request, an internal investigation, or an auditor walking the floor in two years.

There is no bulk close. We considered it, then decided against it. Bulk close encourages "select-all, mark resolved, move on" workflows that destroy the trail and let real risk slip through. One card at a time, one reason at a time. If your inbox is full enough that bulk close feels necessary, the right move is a tighter scan cadence or stricter severity thresholds, not faster closure.

Cards live in four statuses: open, in_progress, snoozed, and closed. The default view shows open and in-progress; snoozed items return on their wake date; closed items are excluded from the default view but remain queryable for audit.

The empty state matters. When the queue is clear, the inbox renders an "Inbox Zero" panel; that is the goal-state of the day. Some teams take a screenshot of the empty inbox at the close of business as a reporting artifact for their HR steering committee.

Here is the routine we recommend to teams onboarding to Clarus HR. It takes 30 to 45 minutes most mornings.

1. 8:45. Open the inbox, sort by severity descending.
2. 8:50. Triage CRITICAL items first. Read each summary, follow the recommended_action link, finish the work, and close with a reason.
3. 9:15. Move to WARNING items. Some are time-boxed (expiring evidence, pending acknowledgments); finish those before moving to non-time-boxed warnings.
4. 9:30. NOTICE and INFO. Most can be closed quickly with a one-line reason. Snooze the few that depend on a Friday payroll close or another scheduled task.
5. End of day. Aim for an empty queue. If items remain, decide whether to escalate, snooze, or carry into tomorrow.

The same data is available via the public API. GET /v1/hr/inbox returns the full open queue with severity, type, summary, recommended_action, and affected_employee_ids. Partners can poll this endpoint and post a digest into a Slack channel each morning, or send a daily email to a manager covering only their team.

Webhook delivery for inbox events lands at General Availability. Until then, the polling pattern is the right call: low risk, easy to debug, and trivially auditable.