Haulmont Introduces AI-Driven Approach to Open-Source BPM

The production team at Haulmont has unveiled a new methodology for developing process-oriented applications using its Russian open-source BPM platform, OpenBPM. This approach emphasizes dynamic, context-aware process diagrams that integrate business intent verification with role definitions, integration requirements, resource allocation, risk assessments, timelines, and financial metrics. The system replaces traditional static BPM diagrams with interactive frameworks that adapt to real-time operational needs.

The methodology includes five distinct phases: establishing enterprise context, conducting user interviews, validating requirements, and ensuring technical feasibility. Knowledge management remains central to the system, requiring dynamic gateways to evolving data sources across multiple formats and storage devices. The platform prioritizes interpretability and relevance of information through caching and canonicalization strategies, while maintaining strict access controls to prevent data leakage.

AI agents play a critical role in this framework by managing knowledge base interactions, resolving conflicting facts, and generating test scenarios without exposing sensitive information. These agents also handle non-functional requirements and document unresolved constraints in development specifications. Interactive sessions with users produce executable artifacts that form the foundation for process applications, with users retaining editing capabilities for precision.

Integration challenges are addressed through mock external task loops that maintain data format specifications while avoiding incomplete system designs. The platform includes sandbox testing environments with preloaded case studies, allowing operators to modify test data without repeated login sessions. Event journals provide transparency by recording decision rationale during process simulations, ensuring multiple test iterations cover both typical and edge business scenarios.

The system emphasizes role-based access controls, treating AI agents as organizational participants with defined permissions. Compliance officers, for example, maintain decision logic authority in credit workflows even without direct task execution. This approach ensures all stakeholders remain visible in process definitions, preventing critical roles from being overlooked during automation.