Why wouldn't I have a team build this in-house?

You certainly could. With enough engineers, bandwidth, and time, anything conceivably can be done. That said, most focus on "the building" part, and not what comes after.  A manual code-based implementation is thereotically feasible, but it leads to a constant error triage cycle with no scalable way to address errors as they show up. Further, if suddenly you need to change the configuration or mapping rules because of a change in the API, or data schema, you're suddenly in trouble; Or, if you need to quickly audit all your data processes to isolate other downstream bottlenecks, you are stuck with no visibility. Data movement is a specialized function, and for it to work 100% of the time, it requires real expertise and foundational technology. The advantages that come from utilizing a platform to manage advanced data transformation and processing can't be overstated.