This study involved an intervention designed to examine whether employability prospects for STEM students studying mathematics could be improved. It incorporated use of non-routine problem solving in second-year mathematics courses at two New Zealand universities. From a theoretical standpoint, we conceptualised a novel construct called "lateral thinking self-efficacy," which is defined as a learner's confidence in their ability to solve non-routine problems. It relates to the creative thinking ability needed for solving innovative real-life problems in the work place, and hence is pertinent in transfer of mathematical learning to novel domains. The findings suggest that the attitude profiles of students with high and low "lateral thinking self-efficacy" differ significantly with respect to three dimensions spanning their affective domain. Further, a significant difference between genders with regard to "lateral thinking self-efficacy" was observed, with a significantly greater proportion of confident males than females, although there was no significant association between gender and non-routine problem-solving performance. These results raise questions about equity with regard to employability prospects for females in STEM companies and have implications for the underrepresentation of women in STEM fields.