Test and deploy your agent

Running Simulations and reading Quality reports

In brief: Simulations have not shipped yet. Until they do, use Preview for ad-hoc checks, Quality reports to score the agent against control questions, and Conversations to spot production answers that need attention.

Testing the Customer Agent involves checking answers before changes go live and monitoring them afterward. Simulations are intended to automate scripted checks, but the available information does not explain how to run or score them in the product. Here’s what you can do now and what is known about the planned feature.

What to know about Simulations

Simulations are scripted conversations that run automatically against the Customer Agent. They are in Beta and on the platform roadmap, but have not shipped yet.

The planned feature is designed to test specific scenarios, group tests into suites, schedule or run them on demand, compare scores across runs, and trace failures to a Source, Procedure, or Living Context value. Until it is available, you cannot use these capabilities to run Simulations.

What to use in the meantime

Three existing tools cover most of the testing needs that Simulations are intended to address:

• Preview for ad-hoc checks.
• Quality reports to score the agent against control questions.
• Conversations to spot answers that need attention in production.

The available information does not describe how to open, interpret, or act on individual Quality reports, so it cannot provide step-by-step guidance for reading them.

When to test

Testing is especially useful before making changes that could affect existing answers. The stated use cases for Simulations include changes to Living Knowledge, publishing a new component or Moment, and scaling to a new market or language.

For now, use Preview, Quality reports, and Conversations for the checks they support, and look to Simulations for automated scenario testing once the feature ships.