Agentforce Builder versus Testing Center testing
Key Takeaway
Start with single-session manual testing in Agentforce Builder, and scale up to batch tests in Agentforce Testing Center when the agent performs reliably in individual scenarios. When a batch test input fails, copy it into the Agentforce Builder Preview panel to review the path the agent took.
Agentforce gives you two levels of agent testing. In Agentforce Builder, you chat with your agent one session at a time and inspect exactly how it reasoned. In Agentforce Testing Center, you run batch tests across many scenarios at once to find patterns, edge cases and gaps. You start in Agentforce Builder, then scale to Agentforce Testing Center when the agent behaves reliably in single conversations.
Two levels of agent testing
| Agentforce Builder | Agentforce Testing Center | |
|---|---|---|
| Kind of testing | Single-session manual testing | Batch testing |
| How it works | You chat with your agent in the Preview panel as a user would, one session at a time. | You run batch tests across many scenarios at once. |
| Question it answers | Why did the agent respond this way in this conversation? | How does the agent perform across many scenarios? |
| Use it to | Validate behavior and troubleshoot in detail | Find patterns, edge cases and gaps |
Batch testing exists because agents are non-deterministic: the same input can produce different correct responses, incorrect responses or hallucinations. You would have to think up every way a user might phrase a request and then test each one by one in Preview, which would take far too long. Batch testing surfaces patterns and edge cases that manual testing misses.