Magnetoencephalography (MEG) provides a unique combination of spatial and temporal resolution for studying the human brain in action. While MEG measurement technology is mature for rigorous basic research and routine clinical investigations, the richness of MEG data is not fully exploited yet but calls for more advanced analysis approaches. In this talk, I will first cover the neurophysiology underlying MEG signals and the basics of MEG instrumentation. I will then illustrate how artificial intelligence can be used to facilitate the analysis MEG signals, providing examples on error processing, motor imagery and selective auditory attention in the human brain. I will also discuss how such paradigms can be applied in brain–computer interfaces. I will also argue that applying normative modelling and artificial intelligence to MEG signals could help in deriving disease-specific biomarkers.