As the title says: some basic AI functions in the mobile app would be great. 
Maybe by using data generated from the desktop app?
This is something we discuss but it is NOT a trivial matter given the limited resources and how underpowered a mobile device is.
Underpowered? daringfireball.net/linked/2016/ … ndroid-a10
I’d say the difference is getting close to negligible as far as raw computing power goes. RAM of course is a different matter.
In case of DEVONthink both amount of memory and memory/disk bandwidth are more crucial than raw CPU power.
iOS ≠ macOS and macOS > iOS
The measure of a device isn’t mere numbers. The capabilities of iOS are far less than macOS. My 2012 MacBook Pro may have “numbers” on par with a brand new iPad Pro, but it is a FAR more capable and productive device in many ways.
Please also note that AI technology depends heavily on RAM. And contrary to the Mac there’s no virtual memory (means: almost unlimited amount of RAM) on iOS. Actually, if you use just a few megabytes too much you’re jettisoned — one reason for the frequent ‘crashes’ of the first versions of DEVONthink To Go 2.0.