Increasing environmental demands are driving product development towards innovative design concepts that require lightweight and cost-effective components. With respect to lightweight design, adhesive bonding technology is capable of efficiently joining dissimilar materials (e.g. CFRP with aluminium) while achieving a high stiffness-to-weight ratio. A major concern with respect to the long-term performance of bonded structures is fatigue loading. Fatigue is a complex phenomenon influenced by external (e.g., loading conditions, temperature) and internal (e.g., geometry of the joint, stiffness of the adhesive) factors. Since fatigue testing is expensive and time consuming, e.g. testing a bonded joint with 1 million cycles at 10 Hz takes up to 28 hours, models capable of predicting fatigue life by using existing experimental data are desirable. Engineering problems with complexity, such as fatigue, might involve nonlinear relationships between data for which neither analytical nor empirical solutions are known or feasible. When there is no satisfactory empirical model that accurately describes a particular phenomenon, but sufficient data are available, artificial intelligence (AI) techniques can be useful to develop data-driven models. AI is known for its ability to solve large-scale problems by learning from experience, which enables the development of quantitative expressions that successfully capture complex relationships between parameters. Disciplines such as materials science and engineering often work with relatively small datasets; therefore, expertise is particularly important in selecting the most relevant features and ensuring that a given model is not prone to overfitting, i.e. it is able to generalize when exposed to new data. In the specific case of fatigue modelling, the integration of heterogeneous data sources could be an alternative to circumvent the limitations of reduced data sets. This approach allows data-driven models, such as AI, to combine fatigue data from individual data sources (e.g. publications and materials databases) by exploiting the similar behaviour of materials within the same class. In this context, this paper presents an AI-based approach for fatigue design of adhesively bonded joints, ranging from (i) using existing datasets to reduce the amount of data required for design to (ii) predicting fatigue life of joints under operationally relevant conditions.