Rapid discovery of novel molecules with tailored properties is crucial for the chemical, petrochemical, and pharmaceutical industries. Traditional approaches to the discovery of molecules with specific physical properties are hypotheses driven and heavily dependent on human intuition. When the latter fails due to the complexity of the problem, computational high-throughput methods are applied. These methods substitute the inverse optimization problem, which is difficult to solve, with a direct grid-search problem. As a result, designing new molecules with optimized properties becomes very inefficient and computationally intensive. Can we accelerate this process of molecular discovery by combining automation with machine learning tools? We develop a holistic approach to designing organic molecules, where an AI-controlled platform co-operates with human experts. Such a platform generates new hypotheses, test them, and come up with a decision about the next iterations. The platform is modular and can be integrated with external lab and computational capabilities. I will illustrate our developments using examples of process optimization and finding lead candidates.