Sometimes the agent can use the API directly
In the past year, I've built agent workflows directly on top of raw data APIs. In the past, I would have built a scheduled ingestion job to pull that data into a warehouse first.
Good tools and clear tool definitions let the agent figure out how to get the data it needs. For some use cases, that's enough to avoid building a more complex pipeline. Some data can stay less structured than it would have before.
Data engineering still matters. You still need data modeling, transforms, validation, and pipelines. Companies that have already done that work have a useful foundation for agents.
Originally posted on X. Lightly edited for reading here.