AI Call Center Supervisor in a Live Company: A Three-Week Case Study of an Autonomous AI Agent Supervising a Phone Team

This technical report shows how the figures in the article "An AI Supervisor Ran a Call Center Team for Three Weeks. Then Someone Checked Its Weekly Report" were calculated. An AI supervisor is a software agent that does a team lead's daily work: it watches the team's numbers, reviews calls, flags problems, answers the team's questions and reports up. The subject is an autonomous Claude agent that supervised the small remote phone team of a US-facing telemedicine company for three weeks, with a human team supervisor, a department head and company leadership kept in the loop.

In 16 days the agent was started 459 times, about 280 of them to answer people in the team chat. In week three the team's talk time with partners was 33% to 80% above its best week before the agent, depending on whether the human supervisor's own calls are counted; the first two weeks with the agent were worse. Both main operational defects of the period (a call-screening setting on a caller's phone and the dialer connection order) were found by people, not by the agent. The agent decided nothing about people and recommended no dismissals. Of the nine checkable claims in its own weekly report, one held exactly as written, three held only under one way of counting, two were misleading and three could not be reproduced from the raw call records.

The report gives the formulas, assumptions, confounders and limitations (no control group, one team, call classification available for week three only). The company operates under a non-disclosure agreement: the report contains only the published figures and how they were calculated; the underlying data is not published.