Additional 4th point: understand the ethical tail risk, if it exists!
Is there an assumption (explicit or otherwise) that the human process your are automating includes a capability that the general public would consider "ethical"?
In the worst case scenario, where: 1. that capability is removed by the automation 2. there are negative consequences 3. the whole world discovers the consequences,
will your business survive? Can you "blame the vendor?"
e.g. https://www.reuters.com/article/us-amazon-com-jobs-automatio...
Amazon made a smart move in pulling this project -- the public perception risk of selling, or even deploying a product that could be seen as "bias, automated" was too great.