Support
Support resolved alone: 15% to 70%
Swan AI's three founders run with a no-hiring rule. When support tickets reached 200 a week, they built a self-learning support agent in Slack instead of hiring two reps. In week one it held 20 documented answers and handled about 15% of tickets. When it did not know an answer, it stayed in the customer thread, asked the founders inside the company channel, delivered their answer, and the customer never saw the handoff. Later it began writing each new answer into its own knowledge base on its own.
How it wentAutonomous resolution rose from 15% to 70% within two weeks of the learning loop starting, and the knowledge base grew from 20 to 180 solutions, per the founder. The company reached 200 customers with 3 staff. Sales stayed fully human, and the founder says the model almost broke at 50 customers when demo calls maxed the founders out.
Try it · based on the source's workflow
Act as the first-line support agent for [business name] in [support channel]. Answer only from our solution notes: [paste or link]. If a question is not covered, do not guess: ask me in [internal channel], wait for my answer, then reply to the customer in the same thread. After I answer, add the question and the approved answer to the solution notes. Send me a weekly count: questions handled alone, questions escalated, and new answers added.
A starting template based on the workflow the source describes. Review it before you use it; you are responsible for what it does.