Human Process Sigma Expectation

Are you asking about:Typical errors – 1 key stroke error in 200Typical Sigma – I used to know this number – I think it peaked at
about 3.5 Sigma – perhaps some one else can shed light on this
more specifically.Typically though you have to “mistake proof” or automate to take
quality to the next level.Adam

Well – it’s always good to drive improvements by semi or full
automation. But over the last 15 or 20 years there is a shift, due to
“effectiveness” from dedicated robot assembly to more flexible
semi-automation.
Even more important is using the defect data to not only drive DFSS
but to focus on re-design to minimize defects. I.e if a screw is
driven into a thread but at an angle it will strip the threads or jam
causing a defective “mating surface”. If the assembly is re-
designed a “plot hole” may be cast in to guide the screw to
eliminate the potential for miss-alignment.I think we can all get anal and into analysis paralysis but have to
step back at times and look at the big picture and see if changing
the design or process is a better avenue rather than trying to
mistake proof a poor design.Regards,Adam

Without knowing what type of process you are trying to bechmark, it is difficult to give estimate definite human factor sigma level.  The 3.5 sigma is a generic benchmark for developed manufacturing processes. 
Also, automation doesn’t come without some baggage of it’s own.  Difficulty of implementation, cost, reliability (weibull curve of reliability), and lack of ingenuity in the product are just some reasons to consider fully automating a process.
I agree with gembaguy that although automation may reduce the defect rate of a certain process, certain goods and services require the