SAP (Business Intelligence Expansion)

Contribution to AI

Business data in the 1990s lived where it was made: inside transaction tables built for posting invoices and issuing purchase orders, structures that punished anyone who tried to ask an aggregate question of them. The Business Warehouse released in September 1998 answered that by copying operational data overnight into a separate system shaped for analysis rather than for writing. The lasting part was not the architecture, which others had also arrived at, but what shipped inside it. Business Content supplied ready-made extractors, InfoObjects and cubes that already knew what a cost centre, a material movement or a customer was in R/3 terms, so a company did not design its own analytical schema — it adopted one that thousands of other firms were also running. That is how a large share of the world's commercial activity came to be recorded in comparable structures, which is a precondition for learning anything from it. Later moves followed the same logic of putting method where the data already sat: HANA's column-store design from 2010 collapsed the analytical and transactional split the warehouse had created, the 2013 purchase of KXEN put automated model building in the hands of analysts who were not statisticians, and machine learning services from 2017 onward were sold as features of the ledger rather than as tools.

Drafted with AI and edited by hand (claude-opus-5, reviewed 2026-08).