| name | organizational-memory-agentic-business-process |
| description | LLM-based agents for automating business process execution using organizational memory. General-purpose LLMs lack organization-specific knowledge; this work extracts and structures organizational knowledge from human-oriented artifacts to enable reliable agentic process execution. Activation: organizational memory, business process automation, agentic bpm, knowledge extraction, process execution, LLM agents. |
| metadata | {"arxiv_id":"2607.03228","published":"2026-07-03","authors":"Lukas Kirchdorfer, Adrian Rebmann, Christian Warmuth, Timotheus Kampik, Theiss Heilker, Gregor Berg","tags":["organizational-memory","business-process","agentic-automation","knowledge-extraction","llm-agents"]} |
Organizational Memory for Agentic Business Process Execution
Overview
LLM-based agents offer new opportunities for automating business process execution beyond the limits of rule-based systems. However, general-purpose LLMs lack the organization-specific knowledge required for reliable execution, which is typically fragmented across human-oriented artifacts such as process documentation, SOPs, and internal wikis. This work introduces an organizational memory framework to bridge this gap.
Key Problem
Knowledge Gap in Business Process Automation
- General-purpose LLMs don't know organization-specific procedures, policies, and context
- Organizational knowledge is fragmented across documents, wikis, and tacit expertise
- Rule-based systems are rigid and can't handle the flexibility needed in real business processes
- LLM agents need structured access to organizational knowledge to execute processes reliably
Key Innovations
Organizational Memory Architecture
- Extracts and structures knowledge from human-oriented artifacts (SOPs, documentation, wikis)
- Provides agents with organization-specific context for process execution decisions
- Bridges the gap between general LLM capabilities and domain-specific process knowledge
Agentic Business Process Execution
- Moves beyond rule-based automation to flexible, LLM-driven process execution
- Agents can handle exceptions, make context-aware decisions, and adapt to variations
- Organizational memory enables agents to act as if they understand the organization
Methodology
- Knowledge Extraction: Mine organizational artifacts for process-relevant knowledge
- Memory Structuring: Organize extracted knowledge into queryable organizational memory
- Agent Integration: LLM agents query organizational memory during process execution
- Execution Framework: Agents execute business processes with access to organizational context
Implications
- Organizational memory as a key component for enterprise agentic AI deployments
- Enables LLM agents to serve as reliable business process executors, not just assistants
- Framework for converting fragmented organizational knowledge into agent-actionable format
- Bridge between traditional BPM (business process management) and agentic AI
Pitfalls
- Organizational knowledge is often tacit and hard to extract from documents
- Process execution reliability depends on memory quality and completeness
- Business processes may have implicit rules not documented anywhere
- Change management: organizational memory must be updated as processes evolve
- Regulatory compliance requirements may constrain agentic decision-making
Activation Keywords
organizational memory, business process automation, agentic bpm, knowledge extraction, process execution, LLM agents, enterprise agents, SOP automation, organizational knowledge
Paper Reference
arXiv:2607.03228 - "Organizational Memory for Agentic Business Process Execution" (Jul 2026)
