In-house: deploying existing data scientists and resources. You have more control over the results which can beef up accuracy and increase quality. However, this approach is time-consuming and expensive, especially if annotators are hired and trained from scratch. Thus, in-housing favors large companies with extensive resources.
Outsourcing: hiring managed teams with pre-vetted staff and pre-built labelling tools. You have less control over workflows, but this approach is considered an optimum choice for high-level periodic projects. Furthermore, outsourcing platforms can also commit knowledge of the industries as they serve several customers.