Middle-Office Automation Active

Time Anomaly Agent

Four detector groups run across every open placement every cycle. Missing timesheets, consecutive misses, hours variance, rate drift. Drafts SMS + email reminders, escalates unresolved cases to HITL, and asks for a human signature before touching your ATS.

Time Anomaly Agent Plan view in plan_ready phase showing five proposed actions across the A1, A2, B, and C detector groups with confidence, severity, and dollar impact.
Agent Plan view — five proposed actions waiting on a human signature before any reminder is sent.

Missed and drifted timesheets are where the week leaks.

Every Monday, account managers triage the same list by hand: who didn’t submit, who’s on week two of no submission, who’s reporting 2× their usual hours, whose pay rate just changed without a corresponding placement update. Miss one cycle and the invoice goes out wrong — or not at all. The exact category of leak the Risk Alert panel on the Command Center is designed to surface, but now with an agent that can act on it.

Four detector groups. One planning pass. Every open placement.

Each cycle, the agent loads tenant configuration (thresholds, templates, excluded placements), scans your ATS for every open placement, and classifies findings into four groups. Each group gets its own recommended action, confidence score, severity, and dollar-impact estimate.

A1

Missing timesheet

Placement active, pay period closed, no timesheet submitted.

→ Send SMS + email reminder
A2

Consecutive miss

Second or third pay period in a row with no timesheet.

→ Escalate to account manager + HITL
B

Hours variance

Reported hours differ materially from the placement’s rolling average (configurable ±%).

→ Flag for account-manager review
C

Rate drift

Pay or bill rate on the timesheet doesn’t match the active placement rate.

→ Open dispute ticket, optionally block invoice

Plan → Approve → Execute — with a human in the loop.

The Time Anomaly Agent follows the same five-phase pipeline as every other StaffingAgent agent. It never touches production data without a human approving the plan, and every gateway action it takes is reversible for 7 days.

1

Scan

Pull every open placement, compute each detector across the current pay cycle.

2

Classify

Bucket each finding into A1 / A2 / B / C with confidence, severity, and impact.

3

Draft

Compose tenant-configured SMS + email using the Message Templates tab.

4

Approve

Human reviews the plan in AgentPlanView, unchecks anything that shouldn’t run.

5

Execute & log

Typed gateway calls are issued one at a time. Each one is reversible for 7 days; all are permanently audited.

Thresholds, templates, and runtime — configurable without code.

Every knob the Time Anomaly Agent relies on is exposed in the Agent Settings view of the Command Center. Platform defaults ship sensible; every tenant can override any field, and every override is diffed against the default and can be reset in one click.

  • Thresholds — hours variance %, rate drift $, consecutive-miss count for HITL escalation
  • Message templates — SMS + email per detector with variable reference and preview
  • Runtime — schedule, dry-run toggle, concurrency, error handling
  • Integrations — SMS provider, email sender, audit log sink
  • Dirty tracking — sticky save bar, per-field reset to platform default, discard changes

See it running on real data.

The Time Anomaly Agent is bundled with the Transform tier alongside VMS Reconciliation and Risk Alert. Dry-run on every tenant before first live execution.

How agents work inside the Command Center →