A natural kind of recommendation is the placement of an icon or image on a page transmitted to the customer. The image could either correspond to a promotion, or could even offer the customer live agent help (for web sites with call centers). If the customer clicks on the image, the web server could assign a live agent to interact with the customer. These techniques are common in existing personalization approaches, and easily supported in our approach.
It is also possible for the decision engine to generate entire pages or frames. For example, at Bell labs we have used the Vortex engine to choose from a number of parameterized page templates, and then choose values for the parameters. In this way, simple automated conversations with the customer can be performed. Extending this technique to support more complete, highly personalized generation of web sites is an important direction, and will require the creation of a rich development environment, and perhaps another layer of abstraction on top of the current Vortex language.