Conversational Search, Analysis and Validation in SAP Cloud Integration

Event: SAP Inside Track Belgium 2026
Date: September 19, 2026
Speaker: Vadim Klimov
Session: Conversational search, analysis and validation in SAP Cloud Integration

Transcript

1. Conversational Search, Analysis and Validation in SAP Cloud Integration Vadim Klimov SAP Integration Architect SAP Inside Track Belgium September 19, 2026
2. The Problem Question 1: Which iFlows target a service at https://demo.dev and will be affected by its decommissioning? Question 2: Are there any iFlows calling Internetfacing APIs without authentication? Question 3: How many iFlows use an outdated version of the Advanced Event Mesh sender adapter? Question 4: Do any iFlows violate authentication security standards?
3. How much time do you think it will take to answer these questions?
4. The Problem | Time To Answer iFlow count 20-25 iFlow complexity low to medium (mostly linear flow logic, less than 15 steps per iFlow)
5. The Problem | Time To Answer iFlow count 1000 iFlow complexity medium to high (modularization, branching logic, multiple external calls, more than 25 steps per iFlow)
6. SAP Cloud Integration | Analysis of Integrations Intelligent Search and Analysis Tools Governance and Quality Assurance Conversational Search Anomaly Detection Validate integrations against best practices and design guidelines managed via Docs-as-Code. Query and analyze integration definitions and configurations using natural language prompts. Identify anomalies and atypical patterns within integration definitions and configurations. Generative AI Machine Learning Retrieval-Augmented Generation (RAG) Unsupervised Transductive Anomaly Detection
7. Conversational Search, Governance and Quality Assurance
8. Key Components AI assistant or agent of your choice MCP Standalone assistant Native integration Agent-as-a-Tool Retrieval Vector database Ingestion SAP Cloud Integration Integration flow Designtime artifacts Participants, connections, flow steps Resources Stores • Directory layout: Split into multiple files by area of concern Definition References Scripts, schemas, mappings, archives Configuration Externalized parameters configuration • Format: Markdown • File structure: Flexible Metadata API Security artifacts Integration standards • Storage: Git-managed repository, filesystem
9. Key Components Claude Chainlit FastMCP Claude plugin Pinecone Haystack GitHub SAP Cloud Integration
10. Ingestion Pipeline 1 Fetch integration packages metadata 2 Fetch integration flows metadata 3 Fetch integration flows bundles 4 Extract integration flows definitions 5 Fetch integration flows externalized configuration 6 Chunk integration flow definitions per element 7 Render chunks and enrich with natural language text 8 Upsert chunks to vector database (incl. embedding using integrated inference)