McKinsey released a report last year on Intelligent Process Automation:
It seems to be the next wave in automating back-office work. It takes Robotic Process Automation, adds analytics, voice interaction, pattern recognition, and other AI modules to make it easier to use and more adaptable to variations in the task -- as such, it enables much more automation.
For a primer on what RPA is, see this short video:
Basically, you create workflows that operate on top of pieces of software like spreadsheets and email programs.
Now imagine simply telling the IPA system, "Every time I get a sales report email, add the results to the tracking.xls file, and send me an email if the estimated sales for next year are 10% higher than year-to-date for this year."
Something like that is just out of reach of IPA; but that will change. If the program isn't sure exactly what you want, it will ask follow-up questions like, "Do you mean you want to re-compute the linear regression estimates in column 5, and compare to year-to-date figures in column 4?" [Shows a spreadsheet]
When BCIs arrive, these follow-up questions could be brought down to a minimum; and the system could eventually learn from the brain data to ask very few -- and make do without any brain input.