Let me start straight away with answers -
Do we need to install KNIME at each client box for access to data?
No, KNIME from what I learn by their website is a workflow and process authoring tool. For clients to access data, the definitely do not need authoring tool.
How can we block the user from accessing nothing other than views that we have created - no new thing the user can do apart from views / data access that has been created by us?
I would recommend you to browse their product catalog. There seems to be a commercial product named KNIME server - http://www.knime.org/knime-server. This is meant to be the portal through which the user can access data and view reports. This portal also has the ways to configure access for users.
If a commercial solution is not viable for you, you could develop standalone Java application or You could integrate the solution to your own portal using JSP or servlets. The other such ways are discussed in this discussion forum - http://tech.knime.org/forum/knime-users/deployment-of-knime-mining-tool-as-stand-alone-appliaction-in-java.
How would the front end data be rendered - does the client needs to be educated on KNIME ?
I believe this stands automatically answered by the answers for previous questions.
I hope you got fair enough idea of the way BI solutions work. There is an IDE sort of software which is used to generate the model, or the data flow process. This software is never deployed in the client machines. This model is exported (generally) in XML or other proprietary format. The exported model could then be hosted for consumption by clients. Some BI solution have portals, some have their own clients etc.