Enterprise machine learning systems are smaller scope than moonshots and larger scope than POCs or academic studies.
Moonshots are the industry changing projects that use breakthrough new technology to implement radical solutions to huge problems: autonomous driving, simultaneous translation, chatbots that can pass the turing test, etc. The kind of project that a tech giant might spend half a billion dollars on over the course of a decade. Because these projects loom large in our collective consciousness they tend to be the first ones that come to mind when we think about AI and machine learning. Using these moonshots as our mental archetypes can be misleading as they represent only a fraction of the machine learning projects that will be executed.
At the other end of the spectrum are academic studies and enterprise proof-of-concept projects. These projects tend to be executed in a limited time frame by an individual or small team communicating informally. The goal might be to explore some data to discover new causal insights or to illustrate the potential predictive power of some data which could later be leveraged in a follow-up project.
The most interesting projects from our point of view fall between these two extremes. That is the development of Enterprise Production Systems powered by machine learning models. These projects require tools and methods that are distinct from what is needed by moonshots or POC studies. Unlike studies, but similar to moonshots, they are long-lived projects developed by teams. However, they are clearly smaller scope than moonshots, with a much greater cost-consciousness and more need to deliver predictable timely results. They also must work within the existing systems and standards established in the enterprise such as: agile system development methodology, data governance guidelines, lean six sigma business process re-engineering procedures, etc.