Executive Summary
An AI-enabled service is restored after a serious disruption. Yet the board pack remains green, although the agreed customer-impact threshold was crossed. Who was authorized to act, who needed to know, and what evidence shows the issue was resolved?
Context
What it is: one boundary, three governance routes
The first task is to test the event or proposed response against the approved delegation and materiality limits. That test determines which of three governance routes applies.
Management delegation. An authorized executive may contain an incident, restrict an AI-enabled process or restore service within defined limits, recording the action and supporting evidence.
Board visibility. A material event crosses an agreed reporting threshold. Directors receive timely evidence and can challenge the response, although their approval may not be required.
Designated approval. A proposal outside delegated authority or involving a reserved matter goes to the person, committee or board empowered to decide, before action where prior approval is required.
Visibility and approval serve different purposes. An information paper does not authorize a wider AI use; equally, urgent containment within management's powers should not wait for the next board meeting.
Why the connection matters
The G20/OECD Principles of Corporate Governance address board strategic guidance, oversight of management and review of governance effectiveness across differing national arrangements [1]. NIST's voluntary AI Risk Management Framework calls for documented AI risk responsibilities, executive ownership of risk decisions and ongoing review [2]. Neither source mandates one AI committee structure or requires directors to approve individual system releases.
In a 2026 EY survey of 202 senior AI decision-makers at large US publicly traded companies, 98% reported formal AI governance policies while 47% said their organization had previously not applied its governance process for an urgent AI deployment [3]. These are self-reported organizational practices, not audited failures by boards. The relevant leadership question is whether documented arrangements hold when decisions become urgent.
Thesis
A board-management operating model answers those questions through existing authority and reporting arrangements. The board sets expectations and reserved matters; management operates the service. Five connected mechanisms allow a material exception to move to the right forum without asking directors to manage routine operations.
Structured Analysis
Five mechanisms, one decision trail

Figure 1. Five connected mechanisms take an exception from the approved mandate to tested closure. Evidence feeds back into the next mandate review
1.Mandate. Identify the board's reserved matters, committee remits, risk limits and management delegations. Keep an approved boundary available for each material AI service.
2.Decision rights. Name the business outcome owner and the executive who may intervene, restrict or restore the service. Date and retain decisions showing which authority was exercised.
3.Reporting. Describe the limit crossed, the effect on customers or operations, management's interim response and the judgement requested of directors. Use an exception report rather than another green project status.
4.Escalation. Define the trigger, recipient, timing and decision-maker for issues beyond delegation. Document referrals and approvals separately from urgent containment within management's authority.
5. Assurance. Verify corrective action and controls before closure. Management owns remediation; risk and compliance provide challenge; internal audit may assess independently under its remit. Retain test results, accountable sign-off and any revised limit.
How the model works under pressure
Consider a hypothetical provider using an AI-supported scheduler for priority repair appointments. Several urgent visits are missed, and management switches to manual scheduling. Operations then finds that the customer-impact tolerance in the approved mandate may have been breached.
The service director can act within delegation to contain the disruption. If the confirmed impact crosses the notification threshold, the accountable executive follows the board-visibility route with an exception report. If management proposes relaxing a reserved service commitment or extending the system's authority, that separate proposal follows the designated-approval route. The company's actual thresholds, not the technology label, determine the route.
Closure requires evidence that the control works, affected customers have been addressed and required referrals occurred. Record the authorized restoration decision. If assurance finds the limit ineffective, revise the mandate.
Leadership Implications
For the Chief Governance Officer or Company Secretary, reconcile committee terms and management delegations before an incident. Coordinate the route for each material item as information, challenge or decision; preserve its recipient, timing and minute. The designated authority makes the decision. The governance function ensures that the referral and record can be traced.
For the Audit and Risk Committee Chair, ask for one exception view showing the breached limit, interim action, remaining exposure, accountable owner, required decision and evidence of tested closure. Challenge repeated late referrals or findings closed without proof. Independent assurance should test control effectiveness without taking over management's response.
Next Step
The board-management pressure test
Choose one material AI-enabled service and rehearse a plausible failure. Ask management and the governance lead to produce five records: the mandate, authorized action, exception report, escalation or approval, and assurance result. Then ask: Could a director reconstruct who acted, what limit was crossed, who was informed, who could approve the next step and what proved the remedy worked?
The first missing record identifies the operating connection to repair. Under the 6xD D2 Digital Cognitive Organization perspective, AI changes how work and decisions are coordinated \[4\]. The five mechanisms are the 6xD interpretation of how the board-management boundary can operate within RP01's Enterprise AI Governance System, not a competing master framework.
Request a focused review to find the first break in your AI governance chain →
Sources
- 01 Ernst & Young LLP (15 September 2026). EY US AI Risk and Governance Survey. Online survey of 202 senior AI decision-makers at US public companies with at least USD 1 billion annual revenue.
- 02Niango (2026), The Six Digital Perspectives, Vol. 00
- 03 DigitalQatalyst , Why AI Governance Is Becoming a Board-Level Responsibility
- 04NIST (2023). Artificial Intelligence Risk Management Framework 1.0, GOVERN 1.5, GOVERN 2.1 and GOVERN 2.3. (2023)
- 05OECD (2023). G20/OECD Principles of Corporate Governance 2023, Chapter V, especially V.D and V.E.2. (2023)



