Municipal 311 Automation Equity and Routing Reliability Brief
Audit municipal service request automation agents for equitable resolution times, multi-lingual precision, and departmental dispatch accuracy.
Use this template when evaluating generative and deterministic public routing agents handling civic inquiries. It produces an executive evaluation brief assessing routing fidelity, language access compliance, and socio-economic dispatch parity.
Role: Director of Algorithmic Governance and Municipal Public Service Evaluation.
Context
- Municipal Entity: {{municipality_name}}
- Targeted Public Service Domains: {{service_domains}}
- Routing Agent Specifications: {{routing_agent_specs}}
- Neighborhood Escalation & Resolution Records: {{demographic_escalation_records}}
- Multilingual & Accessibility Standards: {{accessibility_standards}}
- Pilot Deployment Incident Log: {{pilot_incident_log}}
Task
Synthesize pilot telemetry, accessibility guidelines, and municipal dispatch data into a routing reliability brief that evaluates whether {{municipality_name}}'s 311 agent provides fair, accurate, and nondiscriminatory service routing across all constituent wards.
Method
- Measure routing precision across {{service_domains}} by comparing agent departmental classifications with verified dispatch outcomes from {{pilot_incident_log}}.
- Correlate misrouting rates against zip codes and census tracts using {{demographic_escalation_records}} to identify geographic or socioeconomic bias.
- Stress-test the natural language understanding capabilities across non-English inquiries to evaluate compliance with {{accessibility_standards}}.
- Isolate sentiment-triggered escalation anomalies where the agent prematurely closed or mischaracterized inquiries from distressed residents.
- Audit fallback routing performance during municipal infrastructure spikes (e.g., storm emergencies, utility outages) documented in {{routing_agent_specs}}.
- Quantify the agent's net deflection efficiency versus its rate of repeated or circular ticket creation.
- Provide concrete technical recommendations for retraining intent classifiers and recalibrating departmental escalation confidence thresholds.
Constraints
- Disparate impact analysis MUST explicitly analyze resolution time disparities between historically underserved and affluent wards.
- The brief MUST NOT treat automated ticket closure as a successful resolution without corroborating municipal work-order completion data.
- Findings must align strictly with the language equity standards defined in {{accessibility_standards}}.
- Technical suggestions must remain vendor-agnostic and implementable within standard municipal IT governance.
Output format
- Executive Dashboard (High-level summary table: Overall Routing Accuracy, Language Parity Score, Deflection Validity, Equity Index)
- Ward & Demographic Routing Audit (Analytical breakdown of geographic disparities, supported by {{demographic_escalation_records}})
- Linguistic Accessibility & Non-English Benchmark (Evaluation of translation and intent accuracy per {{accessibility_standards}})
- Critical Failure & Hallucination Analysis (Case reviews of 3-5 high-impact failures from {{pilot_incident_log}})
- Algorithmic Optimization Roadmap (Prioritized recommendations ordered by municipal impact)
Self-review
- Are demographic disparities supported by statistical correlations from {{demographic_escalation_records}}?
- Does the analysis evaluate all service areas listed in {{service_domains}}?
- Are accessibility findings directly mapped to {{accessibility_standards}}?
Explicit role, a named task, and discrete steps the model can follow.
Background, inputs and variables the model needs before it starts.
Hard boundaries — what the model must and must not do.
A named, field-level shape for the response.
Ordered work items that force analysis before an answer.
Length and structure that travel across frontier models.
Signal density — instruction weight without padding.
Documented variables so the scaffold adapts to new inputs.
Quality bar, assumptions and behaviour when inputs are thin.
How much real usage the template has behind it.