Analysing a contract alongside the transactions it governs has usually meant moving both to a third system. IBM has added Unstructured Data Workflows to Netezza, letting file-based content be processed and queried inside the same analytical environment as relational data.
The capability is available in Private Tech Preview. It lets data teams work with documents, PDFs, application logs, emails, technical manuals and images using the SQL skills and analytical processes they already have.
IBM is careful about scope. The stated objective is not to replace specialised document processing platforms or AI services. It is to cut the number of steps before structured and unstructured information can be analysed together.
Enterprise data warehouses were designed around a single assumption, that business data lives in tables. Customer records, transactions, sales reports, inventory and financial data all followed a structured format, which made SQL the universal language for analytics.
IBM argues that assumption no longer holds. Organisations generate more documents, application logs, PDFs, emails, technical manuals and images than ever, and much of that content carries business context. It typically sits disconnected from the structured data driving enterprise analytics.
The industry response has generally been to add processing engines alongside the warehouse. IBM's position is that this expands functionality and complexity in equal measure.
Pipelines rather than one-off processing
The release applies the medallion architecture pattern to unstructured content. Customers land PDFs and other file-based content in a data lake, then build pipelines that progressively refine it. Text is extracted, cleansed and standardised until the content becomes curated, analytics-ready data that can be joined with warehouse tables.
Because the pipelines are automated, new documents landing in the lake are picked up and refined without manual intervention. IBM describes the result as a repeatable, governed path from raw enterprise content to business-ready output. It runs in the same environment that already manages structured data.
The same pipeline capability covers retrieval-augmented generation. Pipelines can parse documents, split them into chunks, generate embeddings and load those embeddings into a vector store end to end. As new content arrives, the pipeline keeps the knowledge base current.
Netezza has native vector capabilities. IBM says organisations can therefore run production RAG applications on their existing warehouse. No separate stack of ingestion tools, embedding services and vector databases is required.
The company's argument for keeping this in the warehouse is operational rather than technical. Every additional processing engine is another environment to secure, monitor, govern and maintain. Over time, IBM writes, organisations spend as much effort managing their analytics architecture as generating insight from it.
By keeping structured and unstructured workflows on one platform, IBM says Netezza reduces unnecessary data movement. Unstructured content becomes another source of business insight rather than a separate problem to solve elsewhere.
The company also frames the release as a market signal. Organisations are asking whether their analytics platform can support a broader range of workloads without another layer of infrastructure.