CUES AI Impact Assessment (AIA)

A structured template for evaluating intended value, operational design, risk, human oversight, and governance readiness before an AI tool or use case moves into pilot or production.

Assessment templatePre-pilot requirementDecision support
Business impactHuman oversightRisk controls

Executive Summary

The AI Impact Assessment helps CUES evaluate whether a proposed AI use case is appropriate, controlled, and aligned with organizational goals before adoption. It is designed to force clarity around purpose, scope, users, risks, and success measures.

Complete an AIA when: you are introducing a new AI-enabled workflow, materially changing an existing one, or expanding a pilot into broader production use.

Template Sections

SectionQuestions to answer
Business purposeWhat problem is being solved? Why use AI instead of a standard workflow or rule-based process?
Scope and usersWho will use the system? Which teams, audiences, or stakeholders are affected?
Data and inputsWhat information is entered, referenced, generated, retained, or exported?
Outputs and decisionsWhat does the tool produce? Does it inform, recommend, summarize, draft, classify, or automate?
Human oversightWhere does a human review, approve, edit, or override results?
Risk considerationsCould the system create factual errors, biased outputs, privacy exposure, or process confusion?
Controls and safeguardsWhat approvals, access limits, logging, prompt guidance, and retention rules apply?
Measures of successWhat adoption, quality, efficiency, or satisfaction metrics will determine success?

Suggested Impact Ratings

Low impact
Assistive only
Medium impact
Operational influence
High impact
Sensitive decisions

Use low impact for drafting, summarization, and internal productivity aids with strong human review. Use medium or high impact when outputs influence member experience, employee evaluation, compliance posture, or any important decision-making workflow.

Approval Checklist

  • Named business owner and technical owner
  • Clear success measures and pilot boundaries
  • Documented human-in-the-loop checkpoints
  • Completed vendor, privacy, and security review as applicable
  • Inventory record created or updated

FAQ

Can one AIA cover multiple workflows? Only when the workflows are materially similar. Otherwise, use separate assessments so controls stay precise.

Is an AIA the same as a DPIA? No. The AIA addresses overall AI-related impact and governance readiness. A DPIA focuses specifically on privacy and data protection risk.