The minimum requirement for mobile devices will soon include continuous collection of several data types, feeding a layer of Machine Learning (A.I.) in the OS that continuously accretes and refines a model of the device’s user; you.
When the device has enough data across enough time and in enough places and situations (real-life scenarios), it can begin to do realtime modeling of your priorities. “The likelies and their uns” of your day and life; your work and leisure.
All the data collected must be encrypted, and exist only on your device. Likewise the predictive models of common scenarios that the device eventually begins to derive about you, the user. It’s too valuable to transmit.
Continuous modeling of your Constant Next Scenario will self-train, comparing real outcomes to its projections. Eventually, the device will reach a point in its practice of ‘knowing’ you that it wll, occasionally, be able to assess your current situation and determine with confidence whether interrupting you is worthwhile.
Just like a smart, experienced administrative assistant.
The next step will be predictive; anticipatory assistance. But only after a lot of learning about you. Assistance in realtime will feel like predictive assistance.