AI less as automation or even tool, and more as medium for sharing knowledge. Example: teams working together to solve problems, with AI continually learning and sharing best practices amongst them so that everyone is enabled and chiseling the frontier of challenges.
Key challenges: accessibility, accountability, asymmetry. There are short term traps in solving only for priveleged problems, exploiting faith in black box models, or failing to share value among data providers, domain experts, and users.