Transdifferentiation is a complete and stable change in cell identity that serves as an alternative to stem-cell-mediated organ regeneration. In adult mammals, findings of transdifferentiation have been limited to the replenishment of cells lost from preexisting structures, in the presence of a fully developed scaffold and niche¹. Here we show that transdifferentiation of hepatocytes in the mouse liver can build a structure that failed to form in development-the biliary system in a mouse model that mimics the hepatic phenotype of human Alagille syndrome (ALGS)². In these mice, hepatocytes convert into mature cholangiocytes and form bile ducts that are effective in draining bile and persist after the cholestatic liver injury is reversed, consistent with transdifferentiation. These findings redefine hepatocyte plasticity, which appeared to be limited to metaplasia, that is, incomplete and transient biliary differentiation as an adaptation to cell injury, based on previous studies in mice with a fully developed biliary system^3-6. In contrast to bile duct development^7-9, we show that de novo bile duct formation by hepatocyte transdifferentiation is independent of NOTCH signalling. We identify TGFβ signalling as the driver of this compensatory mechanism and show that it is active in some patients with ALGS. Furthermore, we show that TGFβ signalling can be targeted to enhance the formation of the biliary system from hepatocytes, and that the transdifferentiation-inducing signals and remodelling capacity of the bile-duct-deficient liver can be harnessed with transplanted hepatocytes. Our results define the regenerative potential of mammalian transdifferentiation and reveal opportunities for the treatment of ALGS and other cholestatic liver diseases.