For decades, SELECT * FROM WHERE has been the golden rule in the SAP ecosystem. From ABAP-managed reporting tools to CDS views in S/4HANA and SAP BW queries, structured query languages have long been the foundation of enterprise analytics. Even NoSQL alternatives eventually folded and introduced SQL-like interfaces. The message was clear: SQL is irreplaceable.
But here’s a question worth asking: After 50 years of teaching computers to understand human language, why are we still forcing humans to speak “computer”?
The End of an Era: SQL in Decline
As AI takes center stage in SAP landscapes, SQL’s role is being redefined. It will likely continue to power legacy systems, but in AI-first applications and next-generation analytics, its relevance is waning.
Natural language is becoming the new standard for querying data. Instead of crafting long SQL statements or navigating InfoProviders, imagine simply asking:
“Show me all suppliers with delayed payments compared to the previous quarter.”
The system interprets your intent and determines:
Should it read from CDS views or unstructured attachments in the Document Store?
Should it run a semantic vector search using embeddings?
Should it trigger external APIs to enrich or verify the data?
No syntax. No debugging. No performance tuning. Just results.
SAP and the Rise of Natural Language Interfaces
This shift is already visible across the SAP product portfolio:
SAP Joule: SAP’s GenAI assistant is being embedded across products to interpret and act on natural language commands.
SAP Analytics Cloud (SAC): Offers search-to-insight capabilities that transform plain questions into live data visualizations.
SAP HANA Cloud: Expands support for vector embeddings, hybrid search, and future LLM integration for semantic search capabilities.
SAP Build Code and SAP AI Core: Enable developers to build and deploy GenAI-enabled business logic directly integrated with SAP systems.
According to Gartner, by 2026, natural language interfaces will become the default way to interact with enterprise data. SQL will go from a must-have to a nice-to-know.
AI Agents: From Queries to Conversations
Natural language input is just the surface. The real transformation lies in AI agents that:
Understand intent: Map business questions to data logic and structure
Choose strategy: Decide whether to use relational filters, semantic vector search, or hybrid techniques
Orchestrate capabilities: Pull data across APIs, tables, or embedded documents
Return answers: In human-readable formats, visuals, or even automated workflows
In the context of SAP, these agents could:
Pull structured financial data from CDS views
Merge it with unstructured vendor sentiment from external sources
Recommend actions via SAP Fiori or trigger workflows via SAP Business Technology Platform (BTP)
This isn’t theoretical. AI-native vector databases like Milvus are already performing semantic searches in milliseconds:
results = collection.search(query_vector, top_k=10, filter=”is_active == true”)
One line. No JOINs. No query plans. No guesswork.
Why SQL Struggles in the AI-First World
SQL excels in structured, transactional systems. But the world of AI is:
Unstructured: Text, images, PDFs, audio
Semantic: Context matters more than exact match
Dynamic: Requires constant enrichment, interpretation, and feedback
Modern AI-driven use cases in SAP, such as intelligent invoice matching, employee sentiment analysis, and predictive maintenance, rely on embeddings, NLP models, and hybrid search, not just relational rows and columns. Forcing SQL to handle this complexity creates inefficiencies.
SQL + Vector: A Beautiful Idea That Executes Poorly
Desperate to stay relevant, traditional databases are bolting vector capabilities onto SQL. PostgreSQL, for example, added the <-> operator for vector similarity search:
SELECT *
FROM items
ORDER BY embedding <-> query_vector
LIMIT 10;
This looks clever, but it’s fundamentally flawed. You’re forcing vector operations through SQL parsers, query optimizers, and transaction systems designed for a completely different data model.
The performance penalty is brutal:
Parser overhead: Vector queries must pass through SQL syntax validation
Optimizer confusion: Query planners optimized for relational joins struggle with similarity search
Storage inefficiency: Vectors stored as BLOBs require encoding/decoding overhead
Index mismatch: Traditional B-trees and LSM indexes are suboptimal for high-dimensional vector search
Vector-native databases like Milvus or Zilliz Cloud avoid these issues by being purpose-built for semantic, AI-first applications.
Vector vs Relational in SAP Terms
AspectRelational (HANA, BW)AI-Native (Vector DBs, GenAI Tools)
Data ModelStructured fields in tablesHigh-dimensional embeddings, multimodal objectsQuery LogicJoins, filters, aggregatesSimilarity search, semantic matchingInterfaceSQL, SQLScript, MDXNatural Language, Python APIs, Function CallingUse CasesReporting, transactionsSearch, classification, recommendations, RAG
What This Means for SAP Developers and Analysts
This paradigm shift is coming fast. Here’s how to prepare:
Learn to work with AI agents and prompt-based systems rather than focusing only on query languages
Prioritize metadata management, business semantics, and knowledge graphs over pure data modeling
Explore tools like SAP AI Core, SAP Build Code, and SAP Graph to create semantic and cross-system data applications
Rethink your role: from a report builder to a designer of intelligent, intent-driven user experiences
We’re moving from static dashboards to dynamic, conversational insights, where the system proactively helps users find what they need before they even ask.
Final Thought: Don’t Write a Query. Say What You Need.
In the near future, you won’t open a SQL console. You’ll describe your business need in plain language. And if your system doesn’t understand you?
Maybe it’s not your skills that need updating; maybe it’s your data stack.
For decades, SELECT * FROM WHERE has been the golden rule in the SAP ecosystem. From ABAP-managed reporting tools to CDS views in S/4HANA and SAP BW queries, structured query languages have long been the foundation of enterprise analytics. Even NoSQL alternatives eventually folded and introduced SQL-like interfaces. The message was clear: SQL is irreplaceable.But here’s a question worth asking: After 50 years of teaching computers to understand human language, why are we still forcing humans to speak “computer”?The End of an Era: SQL in DeclineAs AI takes center stage in SAP landscapes, SQL’s role is being redefined. It will likely continue to power legacy systems, but in AI-first applications and next-generation analytics, its relevance is waning.Natural language is becoming the new standard for querying data. Instead of crafting long SQL statements or navigating InfoProviders, imagine simply asking:”Show me all suppliers with delayed payments compared to the previous quarter.”The system interprets your intent and determines:Should it read from CDS views or unstructured attachments in the Document Store?Should it run a semantic vector search using embeddings?Should it trigger external APIs to enrich or verify the data?No syntax. No debugging. No performance tuning. Just results. SAP and the Rise of Natural Language InterfacesThis shift is already visible across the SAP product portfolio:SAP Joule: SAP’s GenAI assistant is being embedded across products to interpret and act on natural language commands.SAP Analytics Cloud (SAC): Offers search-to-insight capabilities that transform plain questions into live data visualizations.SAP HANA Cloud: Expands support for vector embeddings, hybrid search, and future LLM integration for semantic search capabilities.SAP Build Code and SAP AI Core: Enable developers to build and deploy GenAI-enabled business logic directly integrated with SAP systems.According to Gartner, by 2026, natural language interfaces will become the default way to interact with enterprise data. SQL will go from a must-have to a nice-to-know.AI Agents: From Queries to ConversationsNatural language input is just the surface. The real transformation lies in AI agents that:Understand intent: Map business questions to data logic and structureChoose strategy: Decide whether to use relational filters, semantic vector search, or hybrid techniquesOrchestrate capabilities: Pull data across APIs, tables, or embedded documentsReturn answers: In human-readable formats, visuals, or even automated workflowsIn the context of SAP, these agents could:Pull structured financial data from CDS viewsMerge it with unstructured vendor sentiment from external sourcesRecommend actions via SAP Fiori or trigger workflows via SAP Business Technology Platform (BTP)This isn’t theoretical. AI-native vector databases like Milvus are already performing semantic searches in milliseconds:results = collection.search(query_vector, top_k=10, filter=”is_active == true”)One line. No JOINs. No query plans. No guesswork.Why SQL Struggles in the AI-First WorldSQL excels in structured, transactional systems. But the world of AI is:Unstructured: Text, images, PDFs, audioSemantic: Context matters more than exact matchDynamic: Requires constant enrichment, interpretation, and feedbackModern AI-driven use cases in SAP, such as intelligent invoice matching, employee sentiment analysis, and predictive maintenance, rely on embeddings, NLP models, and hybrid search, not just relational rows and columns. Forcing SQL to handle this complexity creates inefficiencies.SQL + Vector: A Beautiful Idea That Executes PoorlyDesperate to stay relevant, traditional databases are bolting vector capabilities onto SQL. PostgreSQL, for example, added the <-> operator for vector similarity search:SELECT *
FROM items
ORDER BY embedding <-> query_vector
LIMIT 10;This looks clever, but it’s fundamentally flawed. You’re forcing vector operations through SQL parsers, query optimizers, and transaction systems designed for a completely different data model.The performance penalty is brutal:Parser overhead: Vector queries must pass through SQL syntax validationOptimizer confusion: Query planners optimized for relational joins struggle with similarity searchStorage inefficiency: Vectors stored as BLOBs require encoding/decoding overheadIndex mismatch: Traditional B-trees and LSM indexes are suboptimal for high-dimensional vector searchVector-native databases like Milvus or Zilliz Cloud avoid these issues by being purpose-built for semantic, AI-first applications.Vector vs Relational in SAP TermsAspectRelational (HANA, BW)AI-Native (Vector DBs, GenAI Tools)Data ModelStructured fields in tablesHigh-dimensional embeddings, multimodal objectsQuery LogicJoins, filters, aggregatesSimilarity search, semantic matchingInterfaceSQL, SQLScript, MDXNatural Language, Python APIs, Function CallingUse CasesReporting, transactionsSearch, classification, recommendations, RAGWhat This Means for SAP Developers and AnalystsThis paradigm shift is coming fast. Here’s how to prepare:Learn to work with AI agents and prompt-based systems rather than focusing only on query languagesPrioritize metadata management, business semantics, and knowledge graphs over pure data modelingExplore tools like SAP AI Core, SAP Build Code, and SAP Graph to create semantic and cross-system data applicationsRethink your role: from a report builder to a designer of intelligent, intent-driven user experiencesWe’re moving from static dashboards to dynamic, conversational insights, where the system proactively helps users find what they need before they even ask.Final Thought: Don’t Write a Query. Say What You Need.In the near future, you won’t open a SQL console. You’ll describe your business need in plain language. And if your system doesn’t understand you?Maybe it’s not your skills that need updating; maybe it’s your data stack. Read More Technology Blog Posts by Members articles
#SAP
#SAPTechnologyblog
