AI Can Access SAP Data. It Still Needs Business Context to Deliver Answers

SAP platforms produce large volumes of data that companies want AI to optimize. Systems such as SAP S/4HANA, SAP Ariba, Salesforce, and NetSuite generate data that can be prepared for reporting. The problem is that analytics-ready is not answer-ready.

The way AI uses data to provide answers depends on how the business defines it. ERP and business applications record transactions and process steps without preserving the full context needed to answer operational questions. An AI system can retrieve fields and tables, but it still depends on context to understand how those elements are connected.

Precog’s Automated Semantic Modeling capability addresses that gap by generating business-aligned models from user input and data structure. The goal is to move beyond plausible output to answers that can support real business decisions. The release follows a broader company expansion, including a significant investment from Venture Guides.

How Enterprise Data Can Limit AI

Enterprise data is deeply contextual. Terms such as revenue, margin, churn, inventory movement, or returns often carry definitions that vary by company, process, or system.

Becky Conning, Chief Product Officer at Precog, uses a simple operational scenario to illustrate the issue. A warehouse manager might ask which inventory items have not moved in six months, excluding returns. In practice, answering that requires connecting ERP, warehouse, and returns data, along with the business rules that define movement and returns. Moving those tables into a warehouse does not capture that logic.

This is where AI runs into limits. A model working from raw tables does not know how the business defines key terms or how those definitions connect across systems. Outputs may appear reasonable while missing the assumptions that make them useful.

There is also a practical constraint. Sending large volumes of operational data into models introduces cost, governance, and consistency concerns.

For example, a finance team asking an AI system to explain margin or reconcile revenue needs results that are repeatable and auditable. But processing transactional data through a model can produce different answers across runs, creating risk in reporting.

Data can be prepared and queried, but without context, it cannot consistently support operational decisions. The gap is between analytics-ready and answer-ready.

How Precog Is Addressing the Context Gap

Precog seeks to close that gap by moving context closer to the point where data is prepared. Its Automated Semantic Modeling capability generates business-aligned models from a user’s stated intent, the source system, and the structure of the data.

Instead of moving raw tables first and defining meaning later, the approach connects data preparation to the question being asked. A user might ask to optimize inventory or understand customer churn. The platform uses that input to generate the questions, relationships, and metrics needed to support the use case.

That approach sits on top of Precog’s role as a data integration layer. The platform connects SAP and non-SAP systems and delivers data into environments, such as SAP Datasphere, SAP HANA, Snowflake, Amazon Redshift, and Google BigQuery.

Within SAP, it is listed as a partner solution for connecting data from systems like SAP S/4HANA, SAP Ariba, SAP Concur, and SAP SuccessFactors into SAP data platforms.

Precog’s approach reflects what is required for AI to act on enterprise data. Data must be prepared with enough context for models to align outputs with how the business defines and runs its processes. In SAP environments, that requirement is explicit.

Data can be integrated and modeled, but answers depend on whether business context is preserved. Without that layer, analytics can scale, but quality decision support lags.

What This Means for SAPinsiders

• AI governance moves upstream. If business context is missing during data preparation, governance cannot be fixed at the model layer. Companies will need controls over definitions, lineage, and semantic assumptions before AI outputs reach users.
• Business teams become model shapers. Semantic modeling embeds domain expertise into AI infrastructure. Finance, procurement, supply chain, and operations leaders will increasingly define the logic that determines whether AI outputs are usable.
• Connector breadth is no longer a differentiator. Vendors will be judged on how well they preserve meaning across systems as data moves between them. Context quality determines whether AI outputs can support decisions.