Help with survey analysis

First off, I'm not an ignoramus, but I know little of the dark art of statistics. However, I am willing to learn. My question / problem is as follows:

We are carrying out a simple survey amongst a population of 100 employees (no, really, there are 100!), asking simple questions like "How would you rate your manager - less than average, average, better than average".

Now, say we get a result that says 50% of the employees think their manager is less than average, one of the questions we will be asked when we present the results is "How accurate is that figure?"

So I think that I need to address the following:
1. What is the minimum number of responses from 100 employees to make the survey data valid?
2. How do we determine or describe the "accuracy" of each result?
3. If we break the sample further down into departments, what is the minimum number of responses per department?

(I'm assuming 1 and 3 are the same question, but maybe you stats guys do funny things with numbers!)

I'm not really looking for numerical answers here, more a pointer to the methods I should be using to address these.

Any help appreciated.