The area you described is a fit for a quick AI pilot.
What this means for your business
The process runs frequently, the data is largely digital, and the manual load is big enough for a change to be felt. That is the combination we see where the first step tends to be fast and tangible.
What we don't know yet
Your answers do not show us the data quality from the inside, how often real exceptions occur, or how the process behaves at peak. We would verify that in a first conversation - no promises before that.
Recommended step
We suggest a Quick Look: a short, concrete look at this one process with an engineer who does the implementation themselves. You would leave with a clear go / no-go and the scope of a first step.