You need to plan and execute tangible projects that advance the organization. Knowing data science problem archetypes — "what’s possible" — is critical while translating an ambiguous business problem into a concrete plan.
Once you have a plan, you need to oversee the data scientists working on it and empower them to succeed. You don’t need to be data scientists yourself, but you do need to have a ballpark idea of the difficulty of different tasks to effectively manage the project.
Interpreting data-driven results, despite the air of objectivity, has always been tricky. As data scientists produce results, you will need to evaluate and question them, understand what to trust (i.e., what “smells”) and ultimately to make the best-informed decisions from the data, while also keeping a healthy amount of data skepticism.