The alpha-1,3/4-L-fucosidases (EC 3.2.1.111; GH29) BbAfcB from Bifidobacterium bifidum and CpAfc2 from Clostridium perfringens can catalyse formation of the human milk oligosaccharide (HMO) lacto-N-fucopentaose II (LNFP II) through regioselective transfucosylation of lacto-N-tetraose (LNT) with 3-fucosyllactose (3FL) as donor substrate. The current work exploits structural differences between the two enzymes with the aim of engineering BbAfcB into a more efficient transfucosidase and approaches an understanding of structure-function relations of hydrolytic activity vs. transfucosylation activity in GH29. Replacement of a 23 amino acids long a-helical loop close to the active site of BbAfcB with the corresponding 17-aminoacid alpha-helical loop of CpAfc2 resulted in almost complete abolishment of the hydrolytic activity on 3FL (6000 times lower hydrolytic activity than WT BbAfcB), while the transfucosylation activity was lowered only one order of magnitude. In turn, the loop engineering resulted in an alpha-1,3/4-L-fucosidase with transfucosylation activity reaching molar yields of LNFP II of 39 +/- 2% on 3FL and negligible product hydrolysis. This was almost 3 times higher than the yield obtained with WT BbAfcB (14 +/- 0.3%) and comparable to that obtained with CpAfc2 (50 +/- 8%). The obtained transfucosylation activity may expand the options for HMO production: mixtures of 3FL and LNT could be enriched with LNFP II, while mixtures of 3FL and lacto-N-neotetraose (LNnT) could be enriched with LNFP III.