Book a demo
Topics · GOVERNANCE-AGENTS

Asset management AI agent registry Quarterly Review | Moring

A quarterly review of the asset management AI agent registry re-examines every registered agent against its recorded purpose, owner, permissions, and controls. The cadence catches scope drift, orphaned ownership, and unregistered agents, and it produces the recurring evidence that model risk management and EU AI Act obligations expect.

On this page
  1. Short answer
  2. How it works
  3. FAQs
  4. Related reading
  5. Sources

Short answer

A quarterly review of the asset management AI agent registry re-examines every registered agent against its recorded purpose, owner, permissions, and controls. The cadence catches scope drift, orphaned ownership, and unregistered agents, and it produces the recurring evidence that model risk management and EU AI Act obligations expect.

How it works

A registry is only as trustworthy as its freshness, and asset management estates change quickly. CrossCountry Consulting reported in August 2026 that AI has moved from experimentation to early production across asset management, which means new agents are being added continuously, and existing ones are being reconfigured as models and tools change. A quarterly review gives the registry a recurring correction cycle: every registered agent is re-examined against its record, and every running agent found outside the registry is either registered or removed.

The review agenda should be concrete. For each agent: is the owner still correct, does the recorded purpose match what the agent now does, are the permission grants still justified, are the attached tools and models still the ones approved, and is the control configuration intact. These are the same questions Robeco's September 2026 insight on agentic AI in asset management encourages investors to ask, translated into a recurring operational cadence rather than a one-time governance exercise.

Quarterly review also surfaces drift that no individual team sees. A research agent may have accumulated tool grants during a project that ended months ago; a reporting agent's owner may have changed teams without the registry being updated; an agent may have been cloned for a new use case without any registration at all. Individually these are small, but across an estate they quietly invalidate the assurance that risk and compliance teams rely on. The quarterly cycle is the mechanism that finds them.

Cadence should also align with regulatory expectations. Model risk management under SR 11-7 requires ongoing monitoring of models in use, not just validation at launch; applying the same recurring rhythm to agents keeps the registry consistent with existing governance. For firms operating in European markets, the EU AI Act imposes documentation and oversight duties on deployers, and a dated review record per agent is straightforward evidence that oversight is actually happening.

The review should produce more than a checkbox. Each quarter's findings, agents re-approved, grants retired, owners reassigned, unregistered agents remediated, should be recorded against the registry so the trend is visible. A shrinking list of exceptions over successive quarters demonstrates that agent governance is maturing; a flat or growing list indicates the estate is outrunning its controls.

Finally, quarterly review is where escalation belongs. Agents that fail review twice, or that operate outside the registry despite remediation, should be escalated to the governance committee with a decision: fix, restrict, or decommission. That escalation path gives the registry teeth and keeps production honest.

Frequently asked questions

Why should asset managers review their AI agent registry every quarter?

Agent estates change quickly, and asset management firms are adding and reconfiguring agents as AI moves into production workflows. A quarterly review re-examines each registered agent's owner, purpose, permissions, tools, and controls, and finds agents running outside the registry. This catches scope drift, orphaned ownership, and cloned agents nobody registered. The cadence also matches SR 11-7 expectations for ongoing monitoring and produces dated evidence of oversight that EU AI Act duties and internal audit can rely on.

What does a quarterly registry review actually check?

The review checks, for each agent, that the recorded owner is current, the stated purpose matches present behavior, permission grants remain justified, attached models and tools are the approved ones, and controls are configured as intended. It also reconciles running agents against the registry to detect unregistered deployments. In an asset management context this ensures agents touching research, portfolio, or client reporting data remain within their approved scope, and it records findings so trends are visible across quarters.

How does quarterly review connect to model risk management?

SR 11-7 requires ongoing monitoring of models after validation, not a single approval at launch. Agents extend that obligation because they consume model outputs and act autonomously. Reviewing the registry quarterly applies the same monitoring rhythm to the agent layer, confirming which models each agent uses and whether its scope still matches its approved function. This keeps agent governance inside the existing model risk framework rather than creating a parallel process that competes for attention.

What happens when an agent fails its quarterly review?

An agent that fails review should be remediated and re-checked in the following cycle. If it fails twice, or if it continues running outside the registry despite remediation, the case escalates to the governance committee with a clear decision: fix the record, restrict the agent's scope, or decommission it. This escalation path, drawn from the discipline of recurring model risk reviews, gives the registry enforcement power in an asset management estate where untracked agents can quietly gain access to fiduciary data.

Sources

  1. - federalreserve.gov
  2. - federalreserve.gov
  3. - artificialintelligenceact.eu