The 3D interface will facilitate the learning curve by incetivating engagement. Engaging the user with score systems also have been tested and proved being much valuable. And that should not be connected in anyway to salary or financial gains as it's just an intelectual incentive that even non-employees could contribute. At any pace, all this can only be tested in real case scenarios later on.
Now think of a gaming platform with tutorials to "learn how to play" and allowing client side customization to each user on how to "play the game" or in this case access the whole customer data. Find it fast and easy with a full history. Update it with relevant information. Fix wrong data points. Clean and merge what's bloated. Connect data sets in a 3D environment and in a intuitive way.
Customer data has to be available to everyone at the company at any given time. Always. That kind of data is cheap, so it should always be stored locally as well. Think "Distributed Updates and Replication". It has to be quick and easy to find too - there's a whole lot to learn from google on that. So the main issue here is how to make that possible in a 3D world.
Data security will be included for later concern but it's not an easy matter and won't be developed for the beta. Data cryptography at other hand is easy to implement and does increase security from pure technology perspective to virtually 100%.
Finally, the Artificial Intelligence will have two main parts: the compact brain (CB) and the full brain (FB). The FB will work in a cluster as big as we can get while CB will work independent from it and should be able to run even on a smartphone. One will feed the other on how they should work but not on their specific tasks as FB will have no main task.