Contact Center Agent Operating Guide

This guide is for the head of customer experience and the person who runs the queue. It assumes one agent, one queue, and a list of intents the company is willing to finish without inventing policy.

Agencies running service for a client should be able to hand this, completed, to that client. The client owns the relationship. The agent does not get a quieter set of rules than the people on the floor.

Before the agent joins the queue

• The intent list is signed. Anything not on the list is a defect if the agent attempts it.
• The greeting says the customer is talking to an agent of the company, in the language of the conversation.
• The path to a person is a normal action and has been tried while the queue is long.
• Authentication stays in the center’s existing mechanism. The agent receives a yes or a no for specific intents.
• Card numbers never enter the chat, the prompt, or the log. Payment stays on the page or IVR already in PCI DSS scope.
• Notes are written to the system human agents already use, with an author that shows the automation.
• Recording and retention follow the same rule as the rest of the center, including any notice the jurisdiction already requires.
• Supervisors extend the existing quality sample. New defects include invented policy, skipped authentication, and a refused handoff.
• The workforce plan says what work remains for people, and who listens to the harder contacts.
• A supervisor can remove the agent from the queue the same day.

Intents that stay with a person

• Complaints that do not map to a documented answer.
• Customers who ask for a person, or who are in a vulnerable situation the center already treats with a special path.
• Disputes, retention offers, and any change to coverage or to the commercial terms of the account.
• Anything the system of record cannot confirm. The agent does not infer a status from old mail.