NFHA’s AI Risk/Benefit Evaluation Framework

The AI risk/benefit assessment framework helps organizations evaluate the positive and negative impacts of an AI system. It looks beyond traditional technical and economic considerations and integrates downstream impacts on users, communities, and the environment.

Download the AI Risk/Benefit Framework

The tool comprises sixty-one scoreable questions organized into 8 steps: AI Model Assessment, Stakeholder Assessment, Impact at Data Level, Impact at Algorithmic Level, Impact at Output Level, Fairness Metrics and Mitigation, Advancing Fairness and Consumer Opportunity, and Legal & Compliance.

The questions accumulate into a final score that ranges between 33–100. A high score between 80–100 indicates that the benefits of the system may substantially outweigh the risks. A low score, between 33–50, indicates that the risks could substantially outweigh the benefits.

* Disclaimer: The assessment does not attempt to provide legal advice, business advice, or a guarantee that a system is risk free. It is designed to help organizations visualize the balance between potential risks and benefits of an AI system to improve organizational decision-making.

How to Use the AI Risk/Benefit Evaluation Framework

This risk/benefit assessment is expressed as a score, which organizations can use to decide whether the system is aligned with current goals and priorities, and whether it carries an acceptable level of risk.

1. Scope the AI system. Begin by clearly scoping what AI system is under review.
2. Collect necessary materials. Before scoring, gather the materials needed to answer questions accurately and consistently:
   • Model documentation or system architecture descriptions
   • Data sources, data dictionaries, and data governance policies
   • Fairness, bias, or performance testing results
   • Risk assessments, impact assessments, or audit reports
   • Policies related to privacy, security, governance, and oversight
3. Complete the assessment. Answer all 61 questions.

• Record each answer in the correct column
• Each question is labeled as either a Risk or a Benefit in Column D.
• If the question is a Risk, enter the assigned numeric value in Column L.
• If the question is a Benefit, enter the assigned numeric value in Column M.

Calculate the score.

• Benefit Score (B): Sum of all values in Column M

• Risk Score (R): Sum of all values in Column L

• Apply the scoring formula: (B - R) / 2 + 69.5
  • Minimum total score: (28 - 102) / 2 + 69.5 = 32.5 (rounded to 33)
  • Maximum total score: (94 - 33) / 2 + 69.5 = 100

Evaluate the resulting score.

• 80–100: Benefits Substantially Outweigh Risks
  The system demonstrates strong safeguards, measurable benefits, and effective governance.
• 50–80: Benefits and Risks Are Balanced
  The system provides meaningful value but carries unresolved risks.
• 33–50: Risks May Substantially Outweigh Benefits
  The system presents significant risk relative to its benefits.

Use results to inform action.

• Identify specific risk mitigation actions
• Set conditions for deployment or continued use
• Inform leadership, governance committees, or regulators
• Prioritize investments in fairness testing, documentation, or oversight

Monitor and re-run the assessment.

• At regular intervals (e.g., annually)
• After significant model updates or retraining
• When data sources or use cases change
• When new risks, complaints, or regulatory requirements emerge