We describe two studies into ways in which human mathematicians use mathematical examples in their research. In the first study we bring together theoretical and empirical approaches to studying ways in which examples are used in mathematical research, concluding that examples are used for conjecture invention, understanding, plausibility-testing, disproof and modification. Where possible we describe corresponding efforts in automating these aspects of reasoning. In our second study we present an investigation based on grounded theory into example-use during an online math- ematical conversation. These studies suggest ways in which "human-like" example-use in mathematics could be further automated.