What was the impact of the AI solution?

Reduced implementation effort in a pilot migrating legacy control panels to modern web-based interfaces Reduced overall coding effort while preserving system integrity and industrial quality standards Speeded dependency analysis for new features compared with manual review

Source

GCPPilotEvidence: Low40/100

Siemens Knowledge Fabric: graph-based agentic workflows to modernize industrial legacy software

Use case typeCode modernizationUpdated Jun 16, 2026

Siemens and Google Cloud created Knowledge Fabric to help modernize large industrial software codebases and the applications that run on them. The system ingests the software ecosystem into an intelligent agentic workflow that can reason across code, Jira, Confluence, and PDF documentation while preserving explainability and traceability.

Organization: Siemens
Industry: Manufacturing
Location: Germany
Published: June 2026

Reported outcomes

Strategic outcomes

Speed & agilityFaster dependency analysis for new featuresOther strategic outcomeReduced coding effort in a production pilotOther strategic outcomePreserved system integrity and industrial quality standards

Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks

Customer: Siemens
Provider: GCP
Maturity: Pilot
Linked source: Google Cloud Blog

Massive codebases exceeded standard LLM context windows Knowledge was scattered across code, Jira, Confluence, and scanned manuals Legacy industrial systems required explainable and verifiable changes Standard coding assistants lacked the contextual depth needed for modernization Built Knowledge Fabric using knowledge graphs on Spanner Graph, the Google Agent Development Kit, Gemini API, Gemini Enterprise Agent Platform, Gemini CLI, and Anthropic Claude Code Mapped relationships between code and

Customer identity supportedSource describes one deploymentMaturity supported

Primary read

Use case focus

Showing 3 of 3

1Code modernization
2Agent orchestration
3Knowledge management

Massive codebases exceeded standard LLM context windows
Knowledge was scattered across code, Jira, Confluence, and scanned manuals
Legacy industrial systems required explainable and verifiable changes
Standard coding assistants lacked the contextual depth needed for modernization

Built Knowledge Fabric using knowledge graphs on Spanner Graph, the Google Agent Development Kit, Gemini API, Gemini Enterprise Agent Platform, Gemini CLI, and Anthropic Claude Code
Mapped relationships between code and documentation and used graph traversal plus vector search to answer dependency and impact questions
Used specialized agents for deep research, user story drafting, architecture impact analysis, and task breakdown with a human in the loop

Technologies

Spanner Graph Google Agent Development Kit Gemini API Gemini Enterprise Agent Platform Gemini CLI Anthropic Claude Code

Reduced implementation effort in a pilot migrating legacy control panels to modern web-based interfaces
Reduced overall coding effort while preserving system integrity and industrial quality standards
Speeded dependency analysis for new features compared with manual review

Architecture

Knowledge Fabric models code and documentation relationships in Spanner Graph, uses GQL and Spanner ANN embeddings for graph-plus-vector retrieval, and orchestrates multiple specialist agents through the Google Agent Development Kit. The workflow keeps a human in the loop and uses Gemini API / Gemini Enterprise Agent Platform to support reasoning and implementation tasks.

Sources & evidence1

Evidence: Low40/100Evidence strength