G2

Helen G.

Business Partnership Manager

Mid-Market (51-1000 emp.)

10/7/2026

"AI Testing Process for Document Processing"

5/5

What do you like best about Testerly?

They understood that AI testing could not rely on one type of assertion. The team combined normal functional and API checks with context-based validation for AI results, and they paid attention to where a failure happened in the processing flow, not only whether the final output looked wrong. Review collected by and hosted on G2.com.

What do you dislike about Testerly?

Nothing significant comes to mind. Some document formats and edge cases were introduced later by our client, so the test set had to evolve during the project, but Testerly handled that well. Review collected by and hosted on G2.com.

Recommendations to others considering Testerly:

We would recommend involving Testerly early and sharing representative production-like examples from the start. For AI products, agreeing on what can be checked exactly and what needs context-based evaluation makes the QA process much clearer. That approach worked well for our document-processing project. Review collected by and hosted on G2.com.

What problems is Testerly solving and how is that benefiting you?

We needed a consistent way to test an AI product that extracts and validates data from invoices and purchase orders. Testerly introduced reusable datasets and regression scenarios, exact checks for structured fields, context-based validation for less deterministic AI behavior, and clear rules for cases that should be sent for human review. This made release checks more repeatable and failures easier to investigate. Review collected by and hosted on G2.com.