You can make use of the voice recognition system present in Windows 7 for this project. The system can learn to identify different voices. It can be used to give different commands.
However, you idea for identifying the face of the person looks a bit complex. If you want to make a system for identifying people using I.F. sensor then you will need to map the unique signals first from each person in the controller. For this you will need Matlab software. Also, the mapping of the office is also necessary.
What I suggest is that rather than going for such a complex project you can start simple- make a robot which can recognise your voice commands using the software present in Windows 7. The command should specify two parts- one should be message which should be recorded and the other part should be a 'name' which should be mapped to a location where that person sits in the office.
You should skip the part where the robot identifies the person, rather, you can use voice password which should be used by the other person to initiate message delivery by the robot (playing of the recorded message). The robot should not search for the person, rather it should go to his location and wait there for some period till the person returns and receives the message and then it should return to the default position.