Tools for data scientists to start experimenting with AI and Machine Learning (ML) models have improved significantly over the past few years. For example, Automated Machine Learning, or AutoML for short, involves the automatic selection of data preparation, machine learning model, and model parameters for a predictive modeling task. This means that a data scientist can run several different types of predictive models and automatically select the best performing model with usually less than 20 lines of code. How is it then that on the one hand building ML models gets democratized and on the other the vast majority of data science projects never make it into production? It turns out that putting a data science project into production involves a whole lot more than just developing a well-performing model. It requires setting up data pipelines, managing infrastructure, monitoring data drift and model performance, and much more. As a former data scientist at Amazon I have learned what these challenges mean in the business world. And as a Senior Solutions Architect for AI/ML at AWS I now focus on helping our customers to go the journey from initial data science experiment to delivering end-to-end AI solutions in production with tangible business value.