[Wafer: response was truncated before the model finished its internal reasoning. Increase max_tokens, or disable thinking on this model (e.g. chat_template_kwargs.enable_thinking=false), then retry.]

New Orleans is using AI to answer 311 non-emergency calls and is evaluating an AI triage system for 911 calls, a shift designed to reduce the load on human dispatchers. The Orleans Parish Communication District (OPCD) says half of 311 callers receive information the AI was already trained to provide.

The district is now testing an AI Emergency Call Triage system that doesn't handle live crisis responses but analyzes calls and directs them to the right place instantly.

How the system routes calls

Callers are automatically directed to an AI agent that asks whether they are calling about a particular incident. If the system doesn't get the necessary information, key details are automatically transferred to a human dispatcher.

The goal is to cut the volume of emergency calls handled by human operators, giving them more time for calls that require live intervention.

Concerns over bias and accents

According to GoveTech, AI systems can improve turnaround times but can also demonstrate hidden biases based on their training data. If the AI's rigid language models fail to understand callers with stronger accents or distinct voice pitches, the system could lose critical responses.

That risk is especially relevant in New Orleans, where callers come from diverse backgrounds and may be under stress when they call for help.

Why this matters for management professionals

This deployment is a working example of AI automation in high-stakes, customer-facing work. The key lesson for managers: AI can handle routine, repetitive interactions, but the design must include a clear path to human escalation when the system reaches its limits.

OPCD's approach, using AI for triage and routing with humans stepping in when needed, offers a template for organizations weighing similar automation. The 50% automation rate on 311 calls shows measurable workload reduction, but the accent and bias concerns make clear why human oversight stays essential.