liveBook · Manning

Post-market monitoring

As explained in AI Governance

The ongoing collection and review of information about a system after it has been placed on the market or put into service.

Post-market monitoring is the ongoing, planned review of a high-risk AI system after launch. Providers must document a system-specific monitoring plan in the technical documentation; waiting for complaints is not enough. section 7.3.3 section 7.5.1

Make the plan operational by specifying:

  • Which production data and metrics to collect. section 7.5.1
  • How often automated and human reviews occur. section 7.5.1
  • Which changes or thresholds trigger investigation, and what corrective action follows. section 7.5.1

The plan is distinct from the organization-wide quality management system. Providers design it; deployers apply the system’s oversight measures, retain its logs, and report relevant observations. When monitoring finds a problem, providers must respond without delay, potentially restricting, withdrawing, or recalling the system. Serious incidents also trigger a separate reporting duty. section 7.5.1 section 7.3.3

ClaimAssist V5

A monitoring plan could track denial rates by claim category, adjuster override rates, and accuracy against a refreshed test set. Weekly dashboards and monthly human review could identify spikes or drops that trigger investigation, such as suspending automated decisions for an affected category. section 7.5.1

How AI Governance covers it

How deep it goes
Specifies plan contents, review cadence, investigation triggers, and corrective actions without prescribing one universal metric set.
Examples
Uses ClaimAssist V5 metrics for denial rates, adjuster overrides, and accuracy, with dashboards and human review.
What it assumes
Assumes a high-risk system and provider-deployed logging and oversight capabilities; assigns deployers a role in applying controls and reporting observations.
What you can do after
Operate a system-specific monitoring plan with defined data, review schedules, thresholds, and corrective actions.
How it connects
Connects to quality management systems, technical documentation, human oversight, automatic logging, and incident reporting.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of AI Governance. The section links point to the parts of the book it draws on.