The discovery of new optical materials for the next generation of technologies is crucial for economic and societal issues. In particular, such materials are of high importance for the telecommunications, medicine, and low energy consumption lighting. The trial-and-error strategies commonly used to identify new materials are, however, time consuming and repetitive since small modifications in the experimental conditions can have drastic effects on their properties. In this project, a new method enabling the accelerated discovery of optical materials using artificial intelligence (AI) will be initiated and developed. High-throughput (HT) decision-making approach using a multi-agent framework will enable to propose efficiently large sets of new syntheses that will be verified by HT experimentation. The feedback (successes / failures) of the decisions will assist the AI in improving its performances. The automatic iteration between decision and experimentation will be expected to run independently of human intervention. The synergy between the collection of experimental data through a robotic platform and its use through a series of data-driven techniques will be initiated by an interdisciplinary team of chemists, spectroscopists, and data scientists. Two chemical systems will be investigated to establish this new methodology: aluminosilicates and hybrid perovskites. For each of these families of compounds, two optical properties that are important in applications ranging from lasers to solid-state lighting will be characterized in HT manner: the photoluminescence and the second harmonic generation. The objective will be to develop and highlight the efficiency of this new AI methodology by demonstrating its ability to outperform the human strategies.