100% agree. I manage two teams of developers - one builds the infrastructure for a back-office (power user) interface - the other builds the infrastructure for an end-user (casual user) interface. While both are web apps, the usage patterns are VERY different.

The power users know all the special shortcuts. Keyboard mappings, tab order, and also app-specific shortcuts (forcing look-ups to behave certain ways, quick key combos to access forms, etc). They value speed of data entry above almost everything.

The casual users know none of these things, don't care, won't remember them if we try to teach them, and place a much higher value on easy navigation and discovery via visual cues and mouse.

Lots of business programs of the sort where people spend much of their day in them should probably look and act more or less like "bad" old DOS keyboard-focused business programs. Instant responses to input, shortcut-driven. You watch someone use one of those who's been at it for even a few weeks, man, they're fast.

A very good touchscreen interface would probably be as good or better for cases in which typing isn't required, except that most touchscreen interfaces are terrible (janky, frequent missed presses, laggy enough that it's often unclear whether a press has been missed or it's just being slow, confusing navigation—see those new McDonalds ordering kiosks for all of the above) so that doesn't seem to be a realistic option unless you put some kind of UX dictator with great taste in charge of the project.

I've yet to see a graphical enterprise app that approaches 1/10 of the interface speed of a mainframe terminal app.

All app developers should be required to sit and watch an expert user interact with a mainframe interface for at least 30 minutes before they're allowed to make arch decisions.

Neither do the buyers. No one cares about the ergonomics of enterprise applications, because the users are forced to use it.