Field reversals are some of the most prominent and commonly known temporal variations of the geomagnetic field. Polarity changes have been observed in seafloor magnetisation patterns, volcanic records, sediment sequences, speleothem records, and have been reported in geodynamo simulations. However, open questions remain concerning the phenomenology and underlying causes of this process. In particular, there is currently no consensus about the temporal scales over which geomagnetic reversals occur. Numerical simulations aimed at understanding Earth’s million-year evolution have predicted a time scale on the order of thousands of years. On the other hand, analysis of a lacustrine sequence in the central Italian Apennines suggests that the most recent geomagnetic reversal (the Matuyama-Brunhes) took place in as short as 13 years, requiring VGP latitudinal changes of the order of 10 degrees/yr [Sagnotti, L. et al. (2015). GJI, 204(2), 798-812]. This extremely short decadal time scale challenges our current understanding of the geodynamo and present-day numerical models. Here we derive fluid flows at the top of Earth’s outer core that optimise either the rate of dipole decay or directional changes local to the Italian Apennines, subject to a minimal number of physical ingredients. Specifically, we neglect the internal dynamics and prescribe a total flow kinetic energy that is consistent with observational bounds. Our optimal flows can drive an instantaneous VGP latitudinal change of at most 5 degrees/yr during the Matuyama-Brunhes transition. Extending the methodology to account for the spatio-temporal evolution of the magnetic field, we find that this solution does not drive a global reversal, but only changes that are highly localised around the Mediterranean and West-African regions. Solutions that optimise the rate of dipole decay suggest that full reversal would take at least 380 years.