In this paper we revisit our spectropolarimetric and velocimetric analysis of the young M dwarf AU Mic based on data collected with SPIRou at the Canada-France-Hawaii telescope, over a monitoring period of 2041 d from 2019 to 2024. The longitudinal magnetic field, the small-scale magnetic field, and the differential temperature of AU Mic, derived from the unpolarized and circularly-polarized spectra, were clearly modulated with the stellar rotation period, with a pattern that evolved over time. The magnetic modeling with Zeeman-Doppler imaging provides a consistent description of the global field of AU Mic that agrees not only with the Least-Squares Deconvolved profiles of the circularly-polarized and unpolarized spectral lines, but also with the small-scale field measurements derived from the broadening of spectral lines, for each of the 11 subsets of the full data. We find that the large-scale field was mostly poloidal, with a dominant dipole component slightly tilted to the rotation axis which decreased from 1.4 to 1.1kG before increasing at the end of the campaign. The average small-scale field followed a similar trend, decreasing from 2.8 to 2.6kG then rising. The long-term magnetic evolution we report for AU Mic suggests that, if cyclic, the cycle period is significantly longer than 6yr. From velocimetric data, we derived improved mass estimates for the two transiting planets, respectively equal to M_b_=6.3^+2.5^-1.8_M{earth} and M_c=11.6^+3.3^-2.7_M{earth}, yielding very contrasting densities of 0.32^+0.13^-0.10_ and 2.9^+1.1^-0.8_g/cm^3^, and a new 90% confidence upper limit of 4.9M{earth} for candidate planet d (period 12.7d) suspected to induce the transit-timing variations of b and c. We also confirm our claim regarding candidate planet e orbiting with a period of 33.11+/-0.06d, albeit with a smaller mass of M_e=21.1^+5.46_-4.3_M_{earth}_.