Background and In the Partial Oral Treatment of Endocarditis (POET) trial, stabilized patients with left-sided infective endocarditis (IE) were Aims randomized to oral step-down antibiotic therapy (PO) or conventional continued intravenous antibiotic treatment (IV), showing non-inferiority after 6 months. In this study, the first guideline-driven clinical implementation of the oral step-down POET regimen was examined. Methods Patients with IE, caused by Staphylococcus aureus, Enterococcus faecalis, Streptococcus spp. or coagulase-negative staphylococci diagnosed between May 2019 and December 2020 were possible candidates for initiation of oral step-down antibiotic therapy, at the discretion of the treating physician. The composite primary outcome in patients finalizing antibiotic treatment consisted of embolic events, unplanned cardiac surgery, relapse of bacteraemia and all-cause mortality within 6 months. Results A total of 562 patients [median age 74 years (IQR, interquartile range, 65–80), 70% males] with IE were possible candidates; PO was given to 240 (43%) patients and IV to 322 (57%) patients. More patients in the IV group had IE caused by S. aureus, or had an intra-cardiac abscess, or a pacemaker and more were surgically treated. The primary outcome occurred in 30 (13%) patients in the PO group and in 59 (18%) patients in the IV group (P = .051); in the PO group, 20 (8%) patients died vs. 46 (14%) patients in the IV group (P = .024). PO-treated patients had a shorter median length of stay [PO 24 days (IQR 17–36) vs. IV 43 days (IQR 32–51), P < .001]. Conclusions After clinical implementation of the POET regimen almost half of the possible candidates with IE received oral step-down antibiotic therapy. Patients in the IV group had more serious risk factors for negative outcomes. At 6-month follow-up, there was a numerically but not statistically significant difference towards a lower incidence of the primary outcome, a lower incidence of all-cause mortality and a reduced length of stay in the PO group. Due to the observational design of the study, the lower mortality may to some extent reflect selection bias and unmeasured confounding. Clinical implementation of PO regimens seemed feasible and safe.