The effect of training background of persons performing artificial insemination (AI) (herd-owner inseminators (OWNER), AI technicians (AI-T), and fertility consultants (FC)) on pregnancy rate and their ability to detect cows not in oestrus were studied. A total of 1584 re-AI occasions on 754 dairy farms were included. Milk samples for progesterone (P4) analysis in all cases were collected, as were data on the herd, previous breeding attempts, oestrous signs, uterine tone, slipperiness of cervix, and co-operation of the cow. Further breeding attempts and next calving or culling date were sought from registers. The cases were distributed into three categories based on P4 concentrations; <6 nmol/l (no luteal activity, could be in oestrus), 6–10 nmol/l (some luteal activity), and >10 nmol/l (high luteal activity, not in oestrus). Of cows offered for re-AI 7.7% had P4 concentration >10 nmol/l, with no difference between OWNER farms and farms using AI service. OWNERs chose for AI more cows having intermediate P4 than farms using AI service (9.8% vs. 5.9%, p < 0.05). AI-Ts recommended no AI significantly less often than FCs (1.6% vs. 4.9%, p < 0.01). Both groups were equally right: 71% and 68% of cows that were recommended to have no AI had high P4 concentration. Due to courageous and correct rejection of cows with high P4, FCs inseminated proportionally more cows in low P4 and less cows in intermediate P4 than OWNERs (p < 0.05). Of cows finally inseminated, 36.7% became pregnant, with no difference between OWNER farms and farms using AI service. Fertility consultants had higher pregnancy rates than AI-Ts (39.6% vs. 32.6%, p < 0.05). Toneless uterus and sticky cervix at AI significantly correlated with AI occurring at the wrong time (p < 0.001). Behaviour of the cow at AI did not predict P4 concentration. In conclusion, 7.7% of cows offered for re-AI had high P4 concentration. Training of AI personnel increased their ability to detect and reject these cows.