InterSystems, Aegis Trace partner on AI decision traceability

InterSystems and Aegis Trace announced a partnership to improve traceability of AI decisions used by regulated firms. The companies will combine InterSystems’ enterprise data and interoperability platforms with Aegis Trace’s model observability and provenance tools and plan pilot deployments with selected customers before wider availability.

The integration will capture model inputs and outputs, model and version identifiers, runtime metadata, generated explanations and confidence scores. The partners said the system will produce searchable, time-stamped provenance records that compliance teams and auditors can use to reconstruct how specific automated decisions were reached and to link outcomes back to the underlying data and model artifacts.

The solution is designed for both on-premises and cloud IT environments. It will include exportable provenance records and options to connect records with enterprise identity and access controls so audit trails reflect who accessed or modified models and data.

Target industries named by the companies include financial services, healthcare and other regulated sectors that face requirements on explainability, auditability and data lineage. The vendors listed use cases such as responding to regulatory inquiries, conducting post-hoc reviews, investigating incidents, monitoring model drift and supporting validation testing.

An InterSystems spokesperson described the work in a statement as an effort to “preserve that evidence in a way that aligns with existing data governance and audit practices.” Aegis Trace said capturing inputs, outputs and metadata at scale will provide regulated organizations with a practical foundation for model governance and that it will support standards for logs and metadata so records can be consumed by governance platforms and internal audit workflows.

Regulators and industry groups have increased focus on AI governance in recent years. Firms in finance and healthcare face oversight that requires documentation of data handling, version control and explainability, and vendors offering observability and provenance tools have expanded to meet those compliance needs.