Back to Swatches

Civic AI - Casework Triage

Fabric Score 4.1
Civil SocietyPublic ServicesCivicNon-GenAI modelDeveloper ×1

Workflow Diagram

Civic AI - Casework Triage workflow diagram

Fabric Score

ValueEfficacySecurityRiskExternalitiesEfficiency3.85.05.03.34.42.94.1Fabric Score

Task

Assist with intake, triage, categorisation, prioritisation, documentation, and draft support for constituent casework requests.

Intent

Legislators and staff receive a high volume of complex, time-sensitive casework requests across agencies and issue areas. The AI system supports staff by structuring intake, identifying urgency and agency ownership, maintaining a complete case history, and assisting with drafting agency correspondence while preserving constituent privacy and data integrity.

AI Workflow

Input

Casework Identifying AIIncoming casework requests submitted via phone call, web form, email, mail, or in-person interactions.
AI Casework SystemA message from a constituent or an agency regarding a new or ongoing piece of casework.

Process

Casework Identifying AIThe request content is analysed to determine whether it is casework or policy-related.
AI Casework SystemThe message content is analysed to identify the issue area, relevant government agency, urgency level, and required documentation. Personally identifiable information is structured into standardised case fields. Similar cases may be routed to the same internal staff for oversight.

Output

Casework Identifying AIThe message and its content are kept in the policy inbox or rerouted to the casework inbox.
AI Casework SystemDraft communications and status updates prepared for staff review, modification, and approval, with all actions logged within the case history.

Human Oversight Level

Human-Led with AI Assistance

Institutional Oversight Examples

  • •Models must be fully closed, secured, and incapable of leaking data
  • •Casework data is owned by the office
  • •Use is governed by internal office policies, constituent privacy expectations, and applicable federal and state data protection requirements

Risk

Mishandled sensitive information, incorrect agency routing, delay in services provided, or inaccurate case status updates.

Output Modification Telemetry

  • Factual errors / hallucinations25%
  • Unfaithful outputs (policy, tone, or style misalignment)20%
  • Irrelevant information20%
  • Missing or incomplete information15%
  • Other15%
  • Poor usability (unclear language)3%
  • Internal inconsistency2%

Transversal Metrics

Grouped by Fabric dimension.

Efficacy

Accuracy98 %

Efficiency

Modification Rate5 %
Modification Time3 min/output
Verification Time3 min
Rejection Rate2 %
Operational Friction7 hrs/wk
Implementation Overhead1000 hrs
Governance Overhead600 hrs
Time to Launch52 wks

Value

Effort Reduction40 %

Risk

Reliability2 incident/month
Autonomy40 %