You measure how well your agent answers. The yardstick is a test set of questions with expected answers. After every change to the agent or the model, the comparison shows what got better and what got worse. You decide from the figures whether a change stays.
The Agentic Part
An agent is an AI that carries out work steps on its own. Here a testing agent puts the questions of the test set to your agent, compares each answer with the expected one and records the result. After every change to the agent or the model, the same test set runs again. The comparison of the runs shows what got better and what got worse, and nobody has to read each answer by hand.
Main Features
- A test set of questions with expected answers for your agent
- A measurement of how well your agent answers
- The same test set after every change to the agent
- The same test set after every switch of the model
- A comparison that shows what got better and what got worse
Use Cases
- Your customer-service agent gets new instructions. The test set runs again, and the comparison shows whether the answers got better or worse.
- A new model is released, and you want to know whether your agent answers better with it. The same test set answers the question.
What we build for you
- The questions your customers or staff put to the agent
- The expected answers that your specialists define
- Your agent that is to be tested
- The models and changes that are compared
