My wife is a bench scientist. The perspective I get from her is that automation is typically much slower than doing it by hand, especially since there are tools like multi-channel pipettes that give you a lot of the benefits of automation without needing to do any coding. The general task of sucking up liquids is also tough to calibrate due to differences in viscosity. An automation engineer will need to spend a lot of time calibrating while someone who has a lot of experience can go by feel.
This is a very interesting comment because the project I am currently on is trying to assess viscosity with computer vision in basically the same way you are saying.
I am surprised you say that doing stuff is faster by hand, can you elaborate what you wife mean by this? Is the bottleneck the user-friendliness of programming the robots? Because I have a hard time believing the actual motion of the researcher pipetting beats the $500k hamilton liquid handlers... could be wrong though!