When automation goes wrong
The problem with automation of systems is the exceptions to normal function.
If manual invoice entry cost $10 per transaction, and correction of an invoice error cost $15, the occasional error makes little impact.
If online automatic order entry costs $0.10 per transaction, and correction of an order error involves two humans including a manager and costs $100, then a few fat fingered customers could send your business broke.
Ideally, order correction should also be an online process, designed in as part of the system.
When attempting to automate business systems, it helps to first evaluate and refine the existing manual system.
Otherwise: Garbage in -> Garbage out. Just making a mess faster.
AI attempting to learn and replicate the manual system, without insertion of intelligent thought, is a recipe for disastar.
"I really hate this darn machine, I wish that they would sell it. It never does quite what I want, but only what I tell it."