InterSystems-Aegis Trace Collaboration Seeks to Give Clients AI Adoption Confidence

Data integrator InterSystems has engaged Aegis Trace to provide the tools for clients to manage their data in a way that ensures trust in their artificial intelligence-led decision making.

Mounting adoption of AI-powered tools is requiring financial institutions to sharpen their data management capabilities in order to ensure the models’ outputs are safe and trustworthy. Not only that, they must also be able to prove to clients and regulators that the decisions made were based on sound data and analytics.

The partnership announcement was made last month at the InterSystems UKI Data Summit in Birmingham, UK. Data Management Insight spoke to Murilo Silvestre, the founder of Aegis Trace, about the how the arrangement was struck and how the two companies expect it will benefit clients.

Data Management Insight: What difference can this partnership make to financial institutions?

MS: It can help financial institutions adopt AI more confidently while maintaining oversight of decisions that affect customers.

They need less manual effort to reconstruct decisions, and it helps accelerate investigations when an outcome is challenged. It also provides clearer evidence for compliance teams, risk functions, auditors and regulators.

Together, that connects the information behind a decision with the evidence required to examine and explain it later.

DMI: How will clients be able to take advantage of this partnership?

MS: Clients can begin with a specific AI use case, whether it is already in production or being prepared for deployment.

Together, we can identify the information used by the AI workflow, the relevant regulatory and internal controls, and the evidence the institution needs to retain. This creates the basis for a focused proof-of-concept demonstrating how InterSystems’ data capabilities and Aegis Trace’s decision-level evidence can operate within the client’s environment.

Existing InterSystems customers can build on their existing data infrastructure. They can introduce traceability and evidence into selected AI workflows without replacing the systems already supporting their organisation.

DMI: What sort of metrics will clients be given and how should they interpret them?

MS: Reporting is configurable according to the client’s requirements. Through the Aegis Trace dashboard, authorised users can access both current and historical certificate data and examine activity across their AI-supported workflows.

The important distinction is that these metrics are not intended to create a generic “AI trust score”. They allow compliance, risk and business teams to identify exceptions, investigate individual decisions and understand patterns across the organisation.

Leading Challenge

Securing trust in their data and AI-led decision making regularly tops surveys of organisations’ leading technology challenges. Its importance to them was highlighted in an A-Team Group Data Management Insight webinar earlier this year in which expert panellists highlighted how failing to ensure that trust can be costly in terms of not only lost business but error remediation and regulatory censure.

The growth of AI agents has also raised the bar, with Informatica from Salesforce warning of a “trust paradox” emerging as Ai literacy skills lag behind agentic deployment.

A variety of vendors, both new and established, have launched products to help their clients overcome the data trust gap. Among them, GoldenSource has retooled its enterprise data management offering to build a Trusted Contextual Data Layer, which grounds enterprise AI in governed, canonical data. As well, [Quest Software](http://Trusted%20Data%20Management%20Platform) has enhanced its long-standing data migration specialities to address database and metadata management for AI.

DMI: How critical is AI-response validation to business operations and regulatory compliance?

Validation becomes increasingly important as AI moves from assisting employees to influencing or making decisions that affect customers.

A plausible AI response is not necessarily correct, compliant or properly governed. For regulated organisations, the question is less about whether an AI system performs well overall, and more about whether the organisation can demonstrate what happened when an individual outcome is challenged.

The regulatory landscape is also increasingly complex. Our own regulatory intelligence currently maps more than 200 regulatory regimes across 51 jurisdictions relevant to the use of AI in regulated sectors. Requirements can overlap across AI regulation, data protection, consumer protection, conduct, record keeping and sector-specific rules.

Decision-level validation enables organisations to identify exceptions as they occur and retain evidence of the controls applied. Operationally, that supports faster investigation and oversight. From a compliance perspective, it provides a much stronger evidential basis than attempting to reconstruct the decision months or years later.

DMI: How does this partnership ensure trust in AI responses – how are they tested?

The partnership connects the information used by an AI system with evidence of how the resulting decision was assessed and governed.

InterSystems provides the enterprise data infrastructure and retrieval capabilities that help ground AI workflows in information the organisation controls. Aegis Trace then operates at the decision level, capturing the relevant context, assessing the decision against applicable requirements and preserving the resulting evidence.

The objective is not to claim that every AI response is inherently trustworthy. It is to give organisations a repeatable way to test, examine and evidence how an AI-supported outcome was reached and governed.

DMI: How does InterSystems IRIS serve the solution?

MS: Our work with InterSystems includes using InterSystems IRIS as part of the backend architecture supporting the Aegis Trace platform.

InterSystems provides the data infrastructure needed to manage and retrieve enterprise information efficiently while supporting the scale and performance requirements of decision-level traceability.

We are also piloting enhancements to several core workflows to understand where InterSystems IRIS can improve retrieval, processing and integration performance across the Aegis Trace architecture.

DMI: How does the platform account for model drift or prompt-template changes over time?

MS: Traceability is designed into the platform at both the client-model and Aegis Trace level.

Our specialised models and corpus releases are pinned and evaluated against human-reviewed datasets for the relevant workstream before release. System prompts are version-controlled, and assessment behaviour can therefore be compared between releases.

This means that a certificate reflects the system state that existed when the assessment took place. A subsequent model, prompt or regulatory change does not rewrite the historical record.

DMI: How does the solution capture and evidence human intervention?

MS: Aegis Trace captures the AI decision trace together with supporting information supplied by the client.

Where human oversight has taken place, that intervention can be recorded as part of the supporting evidence associated with the decision.

Within the Aegis Trace dashboard, authorised teams can review failed or flagged certificates, add compliance notes and escalate decisions to appropriate legal, compliance or other oversight functions.

Reviews and subsequent actions are recorded as part of the evidence history. An existing certificate is never edited or silently replaced. Any subsequent review or change is appended, preserving the original record and creating a complete chain of evidence showing both the AI-supported decision and the human intervention that followed.