Covestro, a global polymer manufacturer with 48 production sites and approximately 17,500 employees, reduced the time to create master data records from 12 hours to 6 minutes with its MARIS AI agent. According to the AWS for SAP blog post, that is a 99% reduction in cycle time for approximately 1,000 material requests per month, or roughly 12,000 per year.
“Our AI agent MARIS has transformed our approach to material master data maintenance by providing instant, conversational access to complex processes. We are now available to our stakeholders 24/7, with no waiting times and no manual dependencies,” said Sunny Liu, Global Process Owner for Material Master Data at Covestro.
Before MARIS, business users who needed to create new materials, extend a plant assignment, or onboard a new customer or vendor had no direct path into SAP Master Data Governance (MDG). They sent unstructured emails to dedicated MDG experts, who interpreted requests, clarified attributes, corrected errors, checked for duplicates manually and entered the data. A single request could take hours or even days.
MARIS (Material Artificial Realtime Intelligence Specialist) and PARIS (Partner Artificial Realtime Intelligence Specialist) were jointly developed by Covestro and Deloitte Consulting on AWS. MARIS went live in Covestro's legacy SAP ERP (ECC) environment in early 2026 and will transition to SAP S/4HANA, whose go-live is planned for January 2027. PARIS, which handles business partner data with duplicate detection, address validation and golden record enrichment using CDQ technology, is planned to go live exclusively on SAP S/4HANA in early 2027. The post states that PARIS has not yet been deployed in production, so its performance metrics will only be available after go-live.
MARIS is embedded in Covestro's Virtual Assistant (CoVA) and is accessible via browser or Microsoft Teams. It guides users through material creation or extension, validates data against SAP tables in real time, checks for duplicates across multiple systems, enriches requests with derived data, and creates and submits SAP MDG Change Requests directly into the connected approval workflows.
Users interact through a chatbot built on AWS Amplify or through Microsoft Teams, and requests flow through Amazon API Gateway to AWS Lambda, which hosts the agent runtime. A Supervisor Agent, built with the Strands Agents SDK and hosted on Amazon Bedrock AgentCore, classifies user intent and routes it to specialized task agents that run on Anthropic Claude models through Amazon Bedrock. SAP master data is replicated via AWS Glue into Amazon S3 as Data Mesh tables, and Amazon Athena queries them to validate material numbers, plant codes, material types and sales organizations before validated payloads are sent to SAP MDG through SAP BTP Integration Suite.
The post says the pattern is designed to extend to additional master data domains and SAP MDG processes as Covestro completes its ACT S/4HANA transformation.