Reliable Machine Learning

Chapter 14. Practical ML Org Implementation Examples

Organizations are complex entities, and all of their different aspects are connected. Organizational leaders will face new challenges and changes in their organization as a result of adopting ML. To consider these in practice, let’s look at three common organizational adoption structures and how they apply to the organizational design questions we have been considering.

For each of these scenarios, we will describe how the organizational leader has chosen to integrate ML into the organization and the impact of that choice. Overall, we will consider the advantages and likely pitfalls each choice has, but in particular, we’ll consider the way that each choice affects the process, rewards, and people aspects (from the Star Model introduced in Chapter 13). Organizational leaders should be able to see enough details in these implementation scenarios to recognize aspects of their own organizations, and be able to map these into their own organizational circumstances and strategies.

Scenario 1: A New Centralized ML Team

Let’s say that YarnIt decides to incorporate ML into its stack by hiring a single ML expert who develops a model to produce shopping recommendations. The pilot is successful, and sales increase as a result of the launch. The company now needs to make some decisions about how to expand on this success and how (and how much!) to invest in ML. The YarnIt CEO decides to hire a new VP to build and run the ML Center of Excellence ...