Every conversation about artificial intelligence in the enterprise eventually arrives at the same uncomfortable question. Before a model can forecast demand, flag fraud or answer a customer, someone has to ask whether the data feeding it is complete, consistent and correct. In most large companies, the honest answer is “not yet.”
Siva Kavuri has spent the past four years working on that answer. Since 2019 he has published nine research articles on how enterprises move, connect and govern the core business data that sits inside their SAP systems. None of them mention neural networks. All of them deal with the layer that decides whether AI projects deliver value or stall in pilot.
His latest work, released this year, carries that effort into two of the most demanding corners of enterprise data: planning flows that cross ERP boundaries, and warehouse inventory at the level of individual serial numbers.
THE CORE IDEA
AI is only as good as the data underneath it. Kavuri’s research focuses on making that data trustworthy, governed and current, at the point where it is most at risk: when it moves.
THE GROUND HIS RESEARCH COVERS
Taken together, the nine articles map the full life of enterprise data during modernization. Seven problem areas stand out.
Faster starts to S/4HANA. His earliest work looked at how pre-built migration content can be installed and adapted quickly, cutting the weeks that programs typically lose before the first trial load.
Measurable migration quality. He developed an SLA-driven approach for tracking data defects across every test cycle into cutover, turning quality from an impression into a metric.
Hybrid and real-time integration. Two strands of research address how to pull data from several on-premises systems into a single cloud data lake, and how to keep that lake current in near real time using SAP HANA Smart Data Integration on AWS.
Governance of mappings and keys. He has examined how value mappings and cross-reference tables, the translation layers between old and new systems, can be versioned, promoted and verified so that every environment runs on the same rules.
Complex master data relationships. His work on SAP Master Data Governance lays out repeatable patterns for customer and vendor hierarchies and the partner functions that tie ship-to, carrier, invoicing and intercompany relationships together.
Planning integration across ERP boundaries. His February 2023 research studies forecast-to-plan (FTP) integration, where dependent demand must flow reliably in both directions between different ERP platforms during long transition periods.
Serialized warehouse inventory. Published in July 2023, his newest research tackles stock migration in SAP Extended Warehouse Management, including handling units, serialized items and special stock, where a single error can leave physical goods and system records out of step.
WHY THE NEWEST WORK MATTERS
The two 2023 contributions push into areas that are usually treated as specialist territory. Few companies retire an old ERP system overnight. For months or years, planning data has to move between the old platform and the new one. If dependent demand is lost or duplicated in that handoff, production plans and purchase orders go wrong. Kavuri’s integration patterns give architects a structured way to design those flows rather than improvising them.
Warehouse inventory is even less forgiving. Stock in a modern warehouse is nested inside handling units, tracked by serial number and often held under special conditions such as consignment or project stock. Migrating it means preserving every one of those relationships exactly. His strategies for serialization and special stock address a step that can halt shipping on day one if it fails.
Both topics share a theme that runs through all nine articles: the details that are hardest to get right are the ones that matter most once the system goes live.
THE AI CONNECTION
Demand forecasting, inventory optimization and supply chain AI all depend on exactly the data Kavuri studies: planning flows, stock records and master data relationships.
THE IMPACT ON THE FIELD
The link to AI is direct. Machine learning models for demand sensing learn from dependent demand history. Inventory optimization relies on accurate stock positions down to the handling unit. Customer analytics depends on clean hierarchies and correct partner relationships. Each of these is a topic Kavuri has researched. When that data is migrated poorly, models inherit the errors and repeat them at scale.
His work also changes how migration itself is understood. Too often it is treated as a one-time technical task, finished at go-live. Kavuri’s research frames it as the moment when an organization decides how trustworthy its data will be for years afterward. Governed mappings, verified readiness, measured quality and real-time pipelines are not only migration practices. They are the foundations of an AI-ready data estate.
The timing adds to that impact. SAP’s 2027 end of mainstream maintenance for ECC is pushing thousands of companies through S/4HANA programs at the same moment that generative AI has made data quality a board-level topic. Teams facing both pressures need methods they can repeat, not heroics. That is what nine articles of practical, field-tested research now offer.
LOOKING AHEAD
As enterprises move from AI experiments to AI in production, attention will turn from models to the data that feeds them. The questions Kavuri has been answering since 2019, about how data moves, how it is governed and how it stays correct, will sit at the center of that shift. With nine articles behind him and two new ones this year alone, his research is already a reference point for the people building that foundation.
