LA's 911 System Is Failing. Can AI Answer the Call?

Los Angeles has a 911 problem, and it's getting worse.

A March report to the City Council found that LAPD answered just 57.43% of 911 calls within California's 15-second standard in 2024 — far below the 90% benchmark required by state law. The department hired 144 dispatcher trainees in 2024 but only 56 in 2025, while losing 75 operators during roughly the same period. About 100 operators must be on duty across a 24-hour period just to meet minimum staffing requirements.

Into this gap steps Aurelian, a Florida-based public safety technology company that says artificial intelligence can solve the problem. Their AI-powered call taker answers police non-emergency phone lines, takes reports, routes complaints, and handles the routine calls that make up the majority of what overwhelms dispatchers.

The pitch is straightforward: roughly 70% of calls to 911 and non-emergency lines are not emergencies. Barking dog complaints, parking disputes, abandoned vehicles, lost property reports, noise complaints — these don't require a human dispatcher making life-or-death triage decisions. An AI system can handle them, freeing human operators for the calls where seconds matter.

It's an appealing solution. It's also one that raises questions about equity, accountability, and what happens when public safety intersects with artificial intelligence in ways that haven't been tested at scale.

## The Scale of the Problem

Los Angeles isn't the only city struggling with 911 response times, but its numbers are among the most stark:

- **57.43% of 911 calls** answered within 15 seconds (California standard: 90%) - **75 dispatchers lost** in a period when only 56 new hires came in - **100 operators required** per 24-hour period just for minimum staffing - **Growing call volume** driven by population growth, homelessness crises, and increased awareness of non-emergency reporting systems

The result is a system where callers in genuine emergencies face dangerous delays. A heart attack, a car accident, a violent crime in progress — every second a human dispatcher spends taking a report about a barking dog is a second that someone in crisis is on hold.

Aurelian CEO Max Keenan says the company's system can handle roughly 70% of non-emergency calls, from initial intake through report generation, with human oversight. The AI doesn't replace dispatchers for emergency calls — it filters out the noise so dispatchers can focus on the signals.

## How the AI Actually Works

Aurelian's system is designed for the non-emergency side of the equation:

- **Call triage:** When a call comes in on a non-emergency line, the AI answers, identifies the nature of the complaint, and determines whether it requires human attention or can be handled automatically - **Report generation:** For routine complaints (noise violations, parking issues, lost property), the AI takes a statement, categorizes the incident, and generates a report — without a human dispatcher involved in the intake - **Routing:** Cases that require officer follow-up are assigned and dispatched according to existing protocols - **Multi-language support:** The system handles calls in multiple languages, addressing a significant gap in many dispatch centers that struggle with non-English callers - **Escalation:** If the AI detects signs of an emergency during a non-emergency call — a caller mentioning a weapon, sounds of distress, urgent keywords — it immediately transfers to a human dispatcher

The system operates within existing non-emergency lines and doesn't intercept 911 calls directly. But it's designed to reduce the spillover effect: when non-emergency lines are jammed, callers often default to 911, adding to the overload.

## The Case For AI Dispatch

The argument for AI-assisted 911 handling is primarily about arithmetic. Los Angeles cannot hire enough dispatchers to meet its staffing needs. The problem isn't budget alone — it's a nationwide shortage of qualified dispatchers, high turnover rates, the psychological toll of the job, and competition from private sector call centers that pay better for less stressful work.

Against that backdrop, a system that can handle 70% of non-emergency calls isn't replacing humans — it's allowing the humans who are available to focus on the 30% of calls that require judgment, empathy, and split-second decision-making.

There are also potential equity benefits. Multi-language AI support could improve response times for non-English-speaking communities that currently face longer wait times and less accurate call routing. Consistent, standardized intake could reduce the variation in call handling that often disadvantages callers from marginalized communities.

## The Concerns

The case against AI dispatch is about what happens when the system fails — and it will fail, because all systems fail:

**Missed emergencies.** The most serious concern is that an AI system will misclassify a genuine emergency as a routine complaint. A domestic violence victim calling from a non-emergency line to avoid alerting their abuser. A person having a stroke whose speech patterns the AI doesn't recognize as abnormal. A child calling about a parent who's collapsed. These edge cases are where AI triage is most vulnerable.

**Accountability.** When a human dispatcher makes a mistake, there's a person to hold responsible, a training protocol to review, and a clear chain of accountability. When an AI system misclassifies a call, the accountability structure is less clear. Was it a training data issue? A model error? A deployment bug? Who is liable?

**Community trust.** In communities where trust in law enforcement is already fragile, replacing human interaction with an AI system — even for non-emergency calls — could further erode confidence in the system. The perception that "the city doesn't care enough to have a real person answer the phone" is a real political and social risk.

**Data privacy.** Every call handled by an AI system generates data — voice recordings, transcripts, categorizations. That data is valuable, and without clear protections, it could be used for purposes beyond public safety. Who owns it? How long is it retained? Can it be shared with law enforcement for purposes beyond the original call?

**Vendor lock-in.** Aurelian is a private company. If LA (or any city) builds its non-emergency dispatch system around their technology, the city becomes dependent on that vendor for a critical public safety function. What happens if the company raises prices, changes terms, or goes out of business?

## What's Actually Happening

Aurelian's system is still in the proposal and pilot stage for Los Angeles. No contract has been signed, and the city would need to go through a public procurement process before any deployment. Other cities are watching closely — several have expressed interest in similar AI-assisted dispatch systems, and pilot programs are running in smaller jurisdictions.

The technology is not science fiction. Natural language processing has advanced to the point where AI can handle structured, routine conversations with reasonable accuracy. The question isn't whether the technology works — it's whether the safeguards around it are robust enough for a public safety context where failures can be fatal.

## What This Means For You

- **If you live in a city with long 911 wait times, this debate affects you directly.** AI-assisted dispatch is likely coming to your city within the next few years. The question isn't whether to use it but how — and with what guardrails.

- **The 70% claim needs scrutiny.** Handling 70% of non-emergency calls sounds impressive, but the remaining 30% includes the hardest cases — the ones where human judgment matters most. The system's value depends on how well it handles the edge cases, not the routine ones.

- **Escalation protocols are everything.** The difference between a useful AI dispatch system and a dangerous one is how quickly and reliably it escalates to a human when something goes wrong. Demand transparency about escalation rates, response times, and failure modes before your city signs a contract.

- **Multi-language support could be a real improvement.** If implemented well, AI dispatch could significantly improve outcomes for non-English-speaking communities. This is an area where AI has a genuine advantage over understaffed human dispatch centers.

- **Data privacy needs to be addressed upfront.** Before any system is deployed, your city council should establish clear rules about data retention, access, and use. Public safety data is sensitive, and the rules governing it should be set by the public — not by the vendor.

- **This is a staffing problem first and a technology problem second.** AI dispatch can help, but it doesn't fix the underlying issue: cities aren't paying dispatchers enough, supporting their mental health adequately, or making the job attractive enough to retain talent. Technology should augment human capacity, not substitute for the political will to properly fund public safety infrastructure.