Organisations

Classification

Classifications are a High tool for giving interrupts a broad categorization and giving human operators an expecation of an issue they might face. Classifications are also partially used for dictating the order of the queue in the interrupts UI.

Classifications should be manually created in the Classification menu under settings. Each classification should have a unique name and description, a priority and colour can then be assigned to the classification.

We recommend that you give concise yet descriptive descriptions to your classifications if you use classifications in conjunction with Argus auto-classify, this will help Argus accurately auto-classify.

Example Classifications

The right set of classifications depends on what your agents do, but here are some common examples to give you some inspiration:

NameDescriptionPriorityColour
Security ActionThe agent wants to grant permissions, rotate credentials, or modify access controls.HighRed
Destructive OperationAn irreversible action such as deleting production data or user accounts.HighRed
Financial ApprovalRefunds, payments, or changes to pricing, discounts, or contract terms.HighOrange
External CommunicationThe agent wants to send an email, publish a post, or otherwise communicate on behalf of the company.HighOrange
Compliance ReviewDecisions with legal or regulatory exposure, such as content moderation or account bans.HighPurple
Low ConfidenceThe agent's confidence is low or it has encountered conflicting instructions.MediumYellow
Policy GapA novel situation that isn't covered by existing policies or playbooks.MediumBlue
Routine VerificationLow-stakes actions that simply need a human sanity check before proceeding.LowGrey

Because priority partially influences the ordering of the interrupts queue, a human operator working through the inbox will see security and destructive actions first, while routine verifications wait until the higher-stakes work is cleared.

On this page