Users are clever, activity-based design is not
It is important to kid yourself that users are intelligent and want to succeed in using your software. You may have a mountain of evidence that users are lazy dolts, but they don't think so. Live in hope that they aren't dolts. You will avoid avoid insulting your customers. And now for something completely different...
Activity-based design means that the software being designed won't be general purpose. As soon as you make an assumption that a tool will be used for a particular purpose, the design process is skewed toward that use and thus away from all others. The tool will (if the designers have talent) be excellent for the intended use, but poorly adapted to any other purpose.
Historically, runaway successes in software have been unexpected re-applications of software intended for another purpose but built in a general purpose way. For example, web browsers were intended for remote viewing of hyperdocuments. Instead they have been pressed into widespread use as application UI proxies.
An arguable exception would be the spreadsheet: Software Arts intended to build a spreadsheet and that's exactly what they built. I'd argue that it's a UI for arithmetic. That's pretty general purpose if you ask me.
Activity-based design runs counter to generality of purpose. It practically guarantees your application won't be a smash-hit. Its fruits are also quite annoying to use, filled as they are with assumptions about the user's purpose. Go listen to anyone struggling with an Office 2007 product: "Stop that you [creature not born of wedlock]! Where the [reproductive act] have they hidden the [athletically incestuous][feature] now, it was on the toolbar a minute ago, the [accursed] thing keeps changing, I don't know why the [blazes] they had to take the menubar away - noooo I don't want a stupid [reproductive] wizard..."
Activity-based designs are a bit like a fast flowing river. If you want to swim out to sea then you're in luck, it will be quick, easy and straightforward. If you have other plans, get ready to put in a lot of effort resisting the tendencies of the tool. GP tools are more like a still lake: they don't help but they also don't hinder. The user is in charge.
If you ask me, the single most important feature in an adaptive activity driven UI is a way to switch the damn thing off, in favour of a general purpose UI.