There are some potential solutions to deal with each problem. To mitigate biases, banks must ensure that the training data used to develop AI models is diverse and representative of the population it aims to serve. Biases can arise when training data is skewed or does not adequately capture the diversity of individuals and scenarios. Implement robust techniques for detecting and evaluating biases in AI models. This involves regularly auditing models to identify any disparities in outcomes across different demographic groups. Bias detection tools can help assess the fairness of algorithms. Implement fairness constraints or objectives in the learning algorithm. Fairness can be defined in many ways, such as ensuring equal predictive performance across groups or enforcing demographic parity in the outcomes (Ntoutsi et al., 2020). To address the challenge of the black box issue in AI systems, various strategies can be implemented, focusing on post-hoc analysis and corrections. Regular audits represent a fundamental approach, involving periodic examinations of the AI system’s decisions to identify and rectify any biases. Another technique involves outcome-based adjustments, allowing for post-processing modifications if decisions disproportionately impact specific groups. This can be achieved through methods like disparate impact analysis, applied in both pre and post-processing stages (Ntoutsi et al., 2020). Additionally, techniques within machine learning algorithms are employed to tackle biases directly during the learning process. Modifications to the original data distribution include altering class labels near decision boundaries, assigning different weights to instances, and carefully sampling from each group to achieve balance. These interventions aim to address and mitigate biases inherent in the AI system. Techniques designed for explainability, such as LIME (Local Interpretable Model-agnostic Explanations) or SHAP (SHapley Additive explanations), are also recommended for enhancing transparency and interpretability in AI decision-making processes. These techniques generate post-hoc explanations for black-box models,  offering insights into individual predictions (Gramegna & Giudici, 2021). Furthermore, decision accuracy can be improved in a Hybrid Intelligence (HI) system by enhancing collaboration between AI and humans through interpretability and complementarity. Fleisher (2022) emphasizes the need for a balanced approach, integrating human judgment and expertise alongside XAI models. This dual approach ensures fair, ethical, and accurate loan allocation processes, acknowledging the limitations of relying solely on AI systems in contexts demanding transparency and trustworthiness. Lastly, Banks should invest in reskilling and upskilling programs to prepare the workforce for roles that complement AI technologies. Training humans to interpret AI recommendations enhances decision accuracy by actively involving them in the decision-making process (Hemmer et al., 2021). Humans trained to dynamically interact with the AI can better understand the AI’s logic, use their own expertise to evaluate the AI’s recommendations, and decide when it might be appropriate to override the AI’s suggestions. This collaboration between human intuition and AI’s data-driven approach can lead to improved Complementary Team Performance (CTP), contributing to more accurate and reliable decisions in the context of Hybrid Intelligence systems.