AI Agent Builder

The problem

A snapshot of one of our internal operational processes that requires manual tasks and human auditing

Design solution

Custom Evaluations: Write custom evaluation criteria to judge the quality and accuracy of an agent's work

Prompt History: View and restore previous versions of the agent instructions

Training data view: Get a full snapshot of the agent's result and decision-making

My role

Designs shared during office hours were meant to be quick concepts that convey an idea

Outcome