AI Post-Deployment Monitoring Plan

How to use this document

• What it is for. This is a base plan to monitor an AI system once it is in production: what is measured, how often, against which thresholds, who reviews it and what happens when something deviates. It covers performance, data drift, fairness, operations, generative AI behavior, human oversight, complaints, incidents and vendor changes.
• Who should own it. The owner of each system, with its technical team (AI operations or data), coordinated by the AI lead. Enter the function in [RESPONSIBLE FUNCTION] and the person in [OWNER].
• Decisions to take first. Decide whether you will have one plan per system or a common plan with a record per system (Section 2), the monitoring intensity of each system based on its risk level (Section 3), the metrics and thresholds (Section 4) and how long logs will be kept (Section 6).
• Fields to complete. Complete the system record and the metrics catalog (aid B) with specific values. A metric with no threshold and no associated action is useless: remove it or complete it. Keep [OPTIONAL] clauses only if they apply.
• Country variants. This plan has no country variants. If the system may be high-risk, is a medical device or belongs to a regulated industry (financial services, health), keep Section 12 and complete it using the applicable laws listed in "Related regulatory references."
• Review. Have the technical team confirm that the metrics can be measured with the available tools, and have legal and the data protection officer validate the fairness thresholds, industry obligations and the required serious incident reporting step (Section 5.5).
• Approval and publication. Have it approved by the system owner and, for High-risk systems, by [AI COMMITTEE]. Record the approval in "Version history" and share it with the people who operate and oversee the system.

AI Post-Deployment Monitoring Plan

Field	Value
Organization	[LEGAL ENTITY NAME]
Owning function	[RESPONSIBLE FUNCTION]
Plan owner	[OWNER]
System or systems covered	[AI SYSTEM]
Approved by	[APPROVED BY]
Effective date	[DATE]
Plan review frequency	[REVIEW FREQUENCY]

1. Purpose and scope

1.1 This plan sets out how [ORGANIZATION NAME] monitors the operation of its AI systems in production in order to detect performance loss, drift, bias, operational failures, unexpected behavior and unwanted effects on people early, and to act before they cause harm.

1.2 It applies to all AI systems in production listed in the "AI System Inventory," including in-house models, vendor models integrated through APIs and third-party AI tools used in business processes. The intensity of monitoring depends on the system's risk level (Section 3).

1.3 Monitoring starts on the go-live date and continues until the system is retired.

2. Monitored system record

Each system covered has this record:

Field	Content
System and production version	[AI SYSTEM], version [COMPLETE]
Vendor and underlying model	[VENDOR], [COMPLETE]
Purpose and intended use	[COMPLETE]
People affected	[COMPLETE: customers, workers, job applicants, patients, the public]
Risk level and lane (per the "AI System Inventory")	Low (Green) / Medium (Yellow) / High (Red)
Deployment profile (P1 to P5)	[COMPLETE]
Performance baseline (validation results)	[COMPLETE: metrics and reference values]
Groups relevant for fairness analysis	[COMPLETE]
System owner	[OWNER]
Technical owner (system owner's team)	[COMPLETE]
Designated human overseers	[COMPLETE]
Safe state and rollback procedure	[COMPLETE: reference to the "AI Rollback and Safe-State Protocol"]
Monitoring tools and dashboards	[COMPLETE]
Go-live date	[DATE]