Aims: Physical activity improves glycaemic control in type 2 diabetes (T2D), but adherence is low, and diabetes complications are barriers towards adopting physical activity. We investigated adherence and effects of individualized supervised exercise. Methods: Patients with intermediate (level 2) to high (level 3) risk of complications to T2D (stratified by Danish risk stratification model) were offered 12 weeks of exercise. Primary outcomes were working capacity assessed with the Åstrand-Rhyming cycling test (Åstrand), functional capacity assessed with the 30-second chair-stand test (30 s-CST) and health-related quality of life assessed with EuroQoL-5D-5L (EQ-5D-5L). Associations between stratification levels (2 vs 3) and drop-out, changes in 30 s-CST and EQ-5D-5L were analysed using multiple regression. Results: In total 350 patients accepted participation and 254 (73%) completed with 26 patients experiencing an adverse event. Odds ratio (OR) for non-completion were 1.82 [1.02–3.23] (p = 0.043) for patients stratified level 3 vs 2. Improvements were seen in Åstrand 2.71 [1.59; 3.83] ml O₂/kg/min (p < 0.001), in 30 s-CST 2.34 [2.01; 2.67] repetitions (p < 0.001) and in EQ-5D-5L 0.01 (−0.05 to 0.11) (p = 0.002). There were no associations between changes and risk-stratification levels. Conclusions: Clinically relevant changes were seen after exercising regardless of the stratification levels in patients with intermediate to high risk of complications to T2D.