Back to Swatches

USA - Federal Employee Award Writing

Fabric Score 3.1
Civil SocietyPublic ServicesUS GovernmentGenAI modelUser / Operator / Front-line Staff ×1

Workflow Diagram

USA - Federal Employee Award Writing workflow diagram

Fabric Score

ValueEfficacySecurityRiskExternalitiesEfficiency2.52.33.03.53.43.83.1Fabric Score

Task

Write award citations for government employees to recognise exceptional work.

Intent

Save time for supervisors and give more recognition to employees for outstanding work.

AI Workflow

Input

The supervisor writing the award inputs a bio of the employee and recount of events for which the employee is being recommended the award. A completed award may be included as a style/template model.

Process

The supervisor uses the LLM to draft a statement and iterates by following-up with the LLM to revise certain parts or revises manually. Organisations have experimented with sanctioning LLMs for varying sensitivity levels, but the user must ensure their LLM of choice is appropriate for the task.

Output

Supervisor reviews/edits the award before submitting it.

Human Oversight Level

Human-Approved AI

Institutional Oversight Examples

  • •Organizations have specific LLMs for handling sensitive or use-case-specific use-cases
  • •Establishment of agency-level task forces to create GenAI policy within the federal government
  • •Penalty of false information by applicable federal regulations and ethics standards

Risk

Breach of personally identifying information (PII) or sensitive information by using non-approved LLMs that use the data for training future models.

Output Modification Telemetry

  • Factual errors / hallucinations20%
  • Unfaithful outputs (policy, tone, or style misalignment)20%
  • Missing or incomplete information20%
  • Poor usability (unclear language)15%
  • Internal inconsistency15%
  • Irrelevant information10%

Transversal Metrics

Grouped by Fabric dimension.

Efficacy

Accuracy20 %

Efficiency

Modification Rate90 %
Modification Time7 min/output
Verification Time4 min
Rejection Rate20 %
Operational Friction1 hrs/wk
Implementation Overhead0 hrs
Governance Overhead0 hrs
Time to Launch0 wks

Value

Effort Reduction20 %

Risk

Reliability2 incident/month
Autonomy10 %