NGC 5128 is one of the best targets to study AGN-feedback in the local Universe. At 13.5 kpc from the galaxy, optical filaments with recent star formation lie along the radio-jet direction. It is a testbed region for positive feedback (jet-induced star formation). APEX revealed strong CO emission in star-forming regions but also in regions with no detected tracers of star formation. When observed, star formation appears to be inefficient compared to the Kennicutt-Schmidt relation. We used ALMA to map the 12CO(1-0) emission all along the filaments at 1.3"~ 23.8 pc resolution. The CO emission is clumpy and distributed in two main structures: (i) the Horseshoe complex, outside the HI cloud, where gas is mostly excited by shocks and no star formation is observed; (ii) the Vertical filament, at the edge of the HI shell, which is a region of moderate star formation. We identified 140 molecular clouds. A statistical study reveals that they have very similar physical properties that in the inner Milky Way. However, the range of radius available with the present observations does not enable to investigate whether the clouds follow the Larson relation or not. The large virial parameter of the clouds suggests that gravity is not dominant. Finally, the total energy injection in the filaments is of the same order as in the inner part of the Milky Way. The strong CO emission detected in the filaments is an indication that the energy injected by the jet acts positively in the formation of dense molecular gas. The relatively high virial parameter of the molecular clouds suggests that the injected kinetic energy is too strong for star formation to be efficient. This is particularly the case in the Horseshoe complex where the virial parameter is the largest and where strong CO is detected with no associated star formation. This is the first evidence of inefficient AGN positive feedback.