The purpose of this study is to compare the real versus desired statuses of learning processes in Dana, Asia, Alborz and Iran insurance companies. The population under study includes all participants at training courses in the above companies during the year 1388 A.H. (2010). The study utilized a random sample taken from participants with 313 responses required according to Morgan table. Here 305 participants responded the questionnaires.The research was carried out using a descriptive-survey method. The survey instrument for the present study was a researcher-made questionnaire. This questionnaire had four components: need assessment, planning, implementation and evaluation. It was ordered according to Likert scale and finalized by the corresponding professors for the approval of its validity. The reliability coefficient of the questionnaire was 0.93 where descriptive and inferential statistics were used in this study. Data analysis was performed using Statistical Package for the Social Science (SPSS) to test questions and hypotheses. Descriptive statistics included frequency tables, percentages, and various graphs. The inferential statistics included Kolmogorov-smirnov test, one sample t-test, variation analysis, post-hoc Tuckey test and Pearson rho correlation coefficient. The findings of this study at 5% level with 304 d.f. show that there are differences between real and desired states of learning processes in all four insurance companies. However, in Asia insurance company the best learning process with a mean gap of 1.22 was observed with the corresponding values for Dana, Albroz, and Iran insurance companies equal to 1.37, 1.37, and 1.76. Finally after studying and comparing different models of learning processes, ISO 10015 is suggested for this study.