AI Impact Assessment

Technical explanation

The assessment defines the system and its context, identifies affected stakeholders and possible beneficial or adverse impacts, evaluates severity and likelihood, assigns controls and responsibilities, and records residual impacts and monitoring requirements. It is broader than a technical risk assessment because it considers consequences for rights, safety, access, fairness, work, and other stakeholder outcomes.

Business relevance

Impact assessment supports accountable deployment, regulatory readiness, procurement decisions, and defensible governance. It helps decision-makers identify unacceptable uses or missing safeguards before an AI system reaches production.

Implementation example

Before introducing automated candidate screening, an organisation assesses impacts on applicants and recruiters, tests for unequal outcomes, documents human-review and appeal mechanisms, and establishes post-deployment monitoring.

Limitations and common misconceptions

An AI impact assessment is not the same as an AI risk assessment, data protection impact assessment, or algorithmic impact assessment, although they may overlap. A one-time checklist is insufficient when the system, data, users, or operating context changes.

Topics

Sources