Milie was presenting at the Virtual Assistant Summit in San Francisco about the need to design AI around human fears and hopes.
# Yes, we all have weaknesses, but how do you build a Virtual assistant which will assist without hurting feelings?
# Clearly define the roles:
- things that human are better at: negotiate, leading, motivating, …
- things that artificial intelligence is better at: documenting in a structured manner, optimizing schedules, screening large amount of data, …
# How do you capture the relevant labelled dataset when not-documenting is the standard?
- Build a light light version with quick win for the individuals not your dataset
- Care about constraints from decision makers: immediate ROI, no change management, no share of critical data, no business discontinuity
- Lots of training to explain why you are capturing the dataset and how you will be using it.
# CONCLUSION: It is not because you can build it, and that there is a good reason why people should be using your assistant; that individual will use it. Human are humans and they want:
- to be respected: you can’t replace them by machine
- to be important: keep their knowledge
- to protect themselves: not documenting their mistakes
And if you need to redesign your product startegy to fit with the reality of humans; you just have to do it.