When organizations begin experimenting with AI and machine learning, they typically use public cloud services to avoid the cost and complexity of implementing AI-capable systems. As they transition into full-scale adoption of AI, however, they begin to see the drawbacks of the public cloud.