Your team already runs GraphRAG over the product taxono...

The request to reuse the existing GraphRAG pipeline on raw customer-call transcripts ignores the fundamental difference in data volatility and extraction complexity. The current pipeline is likely inefficient because it re-computes the entire extraction process whenever the index is updated.

To support the new, more dynamic corpus, you must pivot the architecture. By implementing a dual-level, incremental indexing strategy, you ensure that only the entities affected by new transcripts are processed. This prevents the system from incurring unnecessary extraction costs and allows the GraphRAG pipeline to scale effectively with the high-volume, unstructured nature of customer-call data.