Most AIDP projects I work on start in the same place: pulling data out of Fusion, CPQ, or Field Service. The Ask-AIDP plugin helps once you’re inside AIDP, but it doesn’t know anything about those source systems. So developers still dig through separate API docs, figure out auth and payloads by trial and error, and rediscover how the objects relate to each other.
I’d like to see the plugin ship with skills covering the Fusion, CPQ, and Field Service REST APIs: endpoints, data models, auth patterns, and the integration approaches Oracle recommends. With that context built in, the code and agents a developer generates would be right the first time much more often. Teams picking up an unfamiliar source system would also get moving faster.
The second ask is documentation. Building pro-code pipelines and agents on AIDP, especially on operational and streaming data, is still new territory for most Oracle shops. Very little published guidance exists. A set of best practices would go a long way, covering:
- Near-real-time ingestion from Oracle applications, change capture, and incremental loads
- When to stream and when to schedule
- How to ground agents in governed data and structure their tools
- Error handling, monitoring, and security for production workloads
- Best Practices on AILK Usage need on when to write back the data and when to not.
Reference architectures for a few common scenarios, such as quote-to-cash, service operations, and field workforce planning, would make this concrete.
Together, these would cut a lot of the ramp-up time on every AIDP implementation built on Oracle applications.
