The effects of twice-daily GLP-1 analogue injections added on continuous subcutaneous insulin infusion (CSII) in patients with poorly controlled type 2 diabetes (T2DM) were unknown. After optimization of blood glucose in the first 3 days by CSII during hospitalization, patients with poorly controlled T2DM were randomized to receive CSII combined with injections of exenatide or placebo for another 3 days. A total of 51 patients (30 in exenatide and 21 in placebo groups) with mean A1C 11% were studied. There was no difference in mean glucose but a significant higher standard deviation of plasma glucose (SDPG) was found in the exenatide group (50.51 ± 2.43 vs. 41.49 ± 3.00 mg/dl, p = 0.027). The improvement of incremental area under the curve (AUC) of glucose and insulinogenic index (Insulin 0-peak/ Glucose 0-peak) in 75 g oral glucose tolerance test was prominent in the exenatide group (p < 0.01). The adiponectin level was significantly increased with exenatide added on (0.39 ± 0.32 vs. -1.62 ± 0.97 μg/mL, in exenatide and placebo groups, respectively, p = 0.045). In conclusion, the add-on of GLP-1 analogue to CSII increased glucose variability and the β - cell response in patients with poorly controlled T2DM.