After the discovery of gravitational waves from binary black holes (BBHs) and binary neutron stars (BNSs) with the LIGO and Virgo detectors, neutron-star black holes (NSBHs) are the natural next class of binary systems to be observed. In this work, we develop a waveform model for aligned-spin NSBHs combining a BBH baseline waveform (available in the effective-one-body approach) with a phenomenological description of tidal effects (extracted from numerical-relativity simulations) and correcting the amplitude during the late inspiral, merger and ringdown to account for the NS tidal disruption. In particular, we calibrate the amplitude corrections using NSBH waveforms obtained with the numerical-relativity spectral Einstein code (SpEC) and the SACRA code. The model was calibrated using simulations with NS masses in the range 1.2-1.4 M ⊙ , tidal deformabilities up to 4200 (for a 1.2 M ⊙ NS), and dimensionless BH spin magnitude up to 0.9. Based on the simulations used and on checking that sensible waveforms are produced, we recommend our model to be employed with a NS mass in the range 1-3 M ⊙ , tidal deformability 0-5000, and (dimensionless) BH spin magnitude up to 0.9. We also validate our model against two new, highly accurate NSBH waveforms with BH spin 0.9 and mass ratios 3 and 4, characterized by tidal disruption, produced with SpEC, and find very good agreement. Furthermore, we compute the unfaithfulness between waveforms from NSBH, BBH, and BNS systems, finding that it will be challenging for the Advanced LIGO-Virgo detector network at design sensitivity to distinguish different source classes. We perform a Bayesian parameter-estimation analysis on a synthetic numerical-relativity signal in zero noise to study parameter biases. Finally, we reanalyze GW170817, with the hypothesis that it is a NSBH. We do not find evidence to distinguish the BNS and NSBH hypotheses; however, the posterior for the mass ratio is shifted to less equal masses under the NSBH hypothesis.