Asset Mgmt AI approval gate Conflict Resolution | Moring
Approval gates in asset management occasionally face conflicts: overlapping policies, competing approvers, or an action permitted under one rule and prohibited under another. Resolution requires a declared precedence order, a decision owner for every gate, and an escalation path so contested actions pause rather than pass silently.
Short answer
Approval gates in asset management occasionally face conflicts: overlapping policies, competing approvers, or an action permitted under one rule and prohibited under another. Resolution requires a declared precedence order, a decision owner for every gate, and an escalation path so contested actions pause rather than pass silently.
How it works
Conflicts inside an approval gate are foreseeable, so they should be designed for rather than improvised. Solytics' governance guidance recommends assigning decision owners to every governance gate so unresolved risks cannot pass silently, and conflict resolution is the mechanism that makes that assignment meaningful when rules disagree.
The most common conflict is policy overlap. An agent's action may be permitted under a research workflow's data rules while brushing against a client-communication restriction, or a budget allowance may conflict with a transaction limit. Resolution starts with a declared precedence order: safety and client-protection rules outrank efficiency rules, hard prohibitions outrank allowances, and the most specific rule outranks the general one.
The second conflict type is competing approvers. Two desks may hold authority over adjacent aspects of one action, as when a data purchase involves both the research budget and vendor-risk review. The gate should route to the stricter reviewer by default, with joint approval configurable for action types the firm designates as crossing ownership boundaries.
The third type is conflicting evidence or judgment, where the rule set passes the action but the assigned reviewer objects. The gate must treat the human objection as controlling, because Robeco's governance research frames accountability as resting with people who remain answerable for what these systems do. An overridden objection should itself require escalation to a higher authority rather than disappearing.
Resolution also needs a stopping rule. When an action cannot be resolved within the configured precedence and ownership structure, it fails closed and enters an exception queue with a service-level expectation, so contested work does not stall invisibly. Arcesium's asset management framework emphasizes defining agent authority precisely to avoid these ambiguities at runtime.
Document the resolution logic where reviewers can see it. Liferay's governance material emphasizes operating models with clear ownership and decision rights, and Adaptive Security's policy template work shows how approved, restricted, and prohibited structures reduce ambiguity before it becomes a conflict. The gate's precedence configuration should be reviewed on the same schedule as the policies it encodes.
Firms that handle conflicts well gain an unexpected benefit: every resolved case sharpens the rule set. Recurring conflicts signal overlapping or stale policies, and the resolution record tells governance teams exactly where the policy text needs repair.
Frequently asked questions
What happens when two policies disagree inside an approval gate?
The gate applies a declared precedence order rather than ad hoc judgment. Safety and client-protection rules outrank efficiency rules, explicit prohibitions outrank allowances, and the more specific rule outranks the general one. If precedence cannot resolve the conflict, the action fails closed and enters an exception queue with a named owner, following solytics-partners' recommendation that decision owners be assigned to every gate so unresolved risks cannot pass silently. The resolution, whichever way it goes, is logged with its rationale so the policy owners can repair the underlying overlap.
How should competing approvers be handled?
Route to the stricter reviewer by default, and configure joint approval for action types that genuinely cross ownership boundaries. In asset management this arises naturally: a data purchase may touch both a research desk's budget and vendor-risk review, and a client-facing draft may involve both compliance and the relationship owner. Robeco's research frames accountability as resting with people answerable for what systems do, so ambiguity about who decides is itself a governance defect. The gate's routing configuration should make the accountable reviewer explicit for every action type in the inventory.
What if the rule set approves an action but the human reviewer objects?
The objection controls. Rules encode prior judgment, but the named reviewer is the accountable decision point, and Robeco's governance questions center on who remains answerable when agentic systems act. When an automated pass conflicts with human judgment, the gate should require escalation to a higher authority rather than allowing the reviewer to be overridden silently, and it should record both positions. Over time these cases are diagnostic: they show where the rule set diverges from informed judgment and deserve the same policy review attention as any other exception pattern.
How do we prevent conflicts from recurring?
Treat recurring conflicts as policy defects rather than operational noise. Each resolution record identifies the overlapping or stale rules that produced it, so governance teams can repair the policy text, adjust tiering, or redraw ownership boundaries. Arcesium's framework stresses defining agent authority explicitly, which prevents many runtime ambiguities before they occur, and Liferay emphasizes operating models with clear decision rights. Review the gate's precedence configuration on the same cadence as the policies it encodes, and retire rules that generate conflicts without adding protection.