While managing high volumes of AppTransactions in QuickBooks, we've noticed that traditional rule-based validation often fails to categorize complex AI-generated invoices correctly.

To improve accuracy, we are looking for better ways to implement llm testing (https://testomat.io/blog/llm-test/) directly within our automation pipelines to verify non-deterministic transaction metadata.

Are there any plans to support external AI testing tools (https://testomat.io/tag/ai-testing/) or APIs that could help us automate the self-healing of these transaction scripts? Providing a bridge for these QA frameworks would drastically reduce manual reconciliation for power users.

Thanks for considering this technical enhancement!