Emergency dispatch centers across the United States are deploying artificial intelligence systems that move beyond answering calls to predicting where emergencies will occur before they happen. The shift from reactive to anticipatory operations is being driven by a convergence of factors: a staffing crisis that has left one in four 911 positions vacant nationwide, call volumes that reached approximately 240 million annually by 2022, and the emergence of AI tools capable of triaging calls, translating languages in real time, and forecasting demand surges using historical data. Centers in Arlington County, Virginia; Jefferson County, Colorado; Charleston County, South Carolina; and cities including San Jose, Portland, and Austin have deployed AI systems that handle nonemergency calls, transcribe conversations, and route resources based on predictive models rather than waiting for the next call to come in.
- Between 2019 and 2022, one in four positions at 911 centers remained vacant; 75% of emergency communications centers lack the budget to increase their workforce, and 82% report difficulty filling open positions
- Jeffcom 911 in Jefferson County, Colorado, reduced nonemergency call volume by up to 39% after deploying an AI-driven chatbot in December 2022, allowing telecommunicators to focus on life-threatening calls
- Arlington County, Virginia, won an AWS Champions Award in 2025 for its AI-powered 911 system using Amazon Connect to manage nonemergency calls including storm damage reports, graffiti complaints, and towing inquiries
- AI-powered predictive analytics now enable dispatch managers to forecast call volume surges using historical data, weather patterns, and local events, allowing proactive staffing adjustments before queues spike
- OSHA’s proposed federal heat standard and NENA’s ongoing AI integration guidelines reflect the growing institutional recognition that dispatch technology must evolve to match the complexity of modern emergency response
The Staffing Crisis That Forced the Technological Shift
The adoption of AI in dispatch centers did not begin as a technology initiative. It began as a workforce crisis. A joint survey by the International Academies of Emergency Dispatch and the National Association of State 9-1-1 Administrators found that between 2019 and 2022, one in four jobs at 911 centers remained unfilled. The National Emergency Number Association reported in 2024 that 75% of emergency communications centers do not have the budget to expand their workforce, and 82% are struggling to fill existing vacancies. A separate survey found that 77% of U.S. first responders wanted their agencies to adopt AI tools.
The math behind the crisis is straightforward. An estimated 80% of daily call volume at 911 centers consists of nonemergency calls: downed trees, broken traffic lights, animal control requests, noise complaints, and information inquiries. Those calls consume the same dispatcher time as cardiac arrests and house fires. When staffing falls short, nonemergency volume creates a bottleneck that delays response to genuine emergencies. Jeffcom 911 in Jefferson County, Colorado, which fields approximately 750,000 calls annually (250,000 emergency and 500,000 nonemergency) and dispatches for 30 agencies, went 22 consecutive months without meeting NENA’s standard that 90% of 911 calls be answered within 15 seconds before deploying an AI chatbot in December 2022.
After the chatbot went live, Jeffcom 911 met the 15-second standard in six of the following twelve months and came close in two others. Deputy Director Michael Brewer described the AI system as a “force multiplier” that allowed telecommunicators to focus on emergency calls. The center reported no complaints from callers handled by the chatbot. The results demonstrated that AI could address the staffing crisis not by replacing dispatchers but by removing the nonemergency volume that was drowning them.
From Call Triage to Predictive Resource Positioning
The current generation of AI tools in dispatch centers operates across several functions simultaneously. Natural language processing enables real-time call transcription, allowing dispatchers to read a caller’s words on screen even when audio quality is poor or the caller is difficult to understand. Real-time translation systems convert calls from non-English speakers into English text instantly, eliminating the delays that previously occurred when dispatchers patched in human interpreters. Charleston County, South Carolina, implemented AI-powered 911 dispatch software in March 2025 that prioritizes calls, transcribes speech, and tracks caller locations, producing faster response times for critical emergencies while reducing dispatcher workload.
The predictive layer represents the next stage. AI systems are now analyzing historical call data, weather patterns, local event schedules, traffic conditions, and time-of-day patterns to forecast where and when call volume will spike. That capability allows dispatch managers to adjust staffing levels and preposition ambulances, fire units, and police resources before incidents occur rather than scrambling to deploy after calls arrive. During large-scale events like severe storms or public gatherings, predictive models can identify neighborhoods with the highest probability of medical emergencies, structure fires, or traffic incidents, enabling resource allocation based on data rather than instinct.
The shift from reactive dispatching to predictive positioning has operational implications that extend beyond call centers. EMS agencies using predictive analytics can reduce average response times by stationing ambulances in locations optimized for the most likely next call rather than returning units to fixed stations after each run. Fire departments can pre-stage equipment in areas where weather data indicates elevated wildfire or structure-fire risk. The approach treats emergency response as a continuous optimization problem rather than a series of independent events.
Arlington County and the Cloud-Based Model
Arlington County, Virginia, has become one of the most widely cited case studies for AI deployment in 911 operations. The county’s Emergency Communications Center implemented a cloud-based system using Amazon Connect that employs AI and automated voice response to manage nonemergency calls. Storm damage reports, graffiti complaints, towing inquiries, and other routine calls are handled by the AI system, freeing human dispatchers to focus on emergencies. The system earned Arlington County an AWS Champions Award in 2025 and has been referenced by NENA and multiple public safety organizations as a model for other jurisdictions.
The Arlington implementation illustrates a broader architectural shift. Legacy 911 systems run on premises-based hardware that limits scalability and makes upgrades expensive. Cloud-based systems can scale dynamically during call surges, integrate AI modules without replacing existing infrastructure, and provide redundancy that protects against facility outages. The transition to Next Generation 911 (NG911), which replaces analog telephone infrastructure with IP-based networks, is accelerating this shift by creating the digital foundation that AI tools require to function.
Cities including San Jose, California; Portland, Oregon; and Austin, Texas, have deployed AI-based virtual agents that handle nonemergency calls, answer routine questions, and gather preliminary information before transferring callers to human operators when escalation is needed. The common thread across these deployments is the same: AI handles volume, humans handle judgment. The technology absorbs the repetitive, time-consuming calls that consume dispatcher capacity, while life-threatening emergencies continue to reach trained human operators who can exercise the empathy, contextual reasoning, and split-second decision-making that AI systems cannot replicate.
The Ethical and Operational Risks
The integration of AI into emergency dispatch raises concerns that the public safety community is actively debating. Algorithmic bias is a primary worry. If predictive models are trained on historical crime or incident data that reflects decades of unequal policing or resource allocation, the system may replicate and reinforce those patterns, directing disproportionate resources to some neighborhoods while underserving others. Researchers at NYU’s Law Review have documented how biased training data in predictive policing systems produces outcomes that mirror the civil rights violations embedded in the data itself.
Accuracy presents another challenge. AI systems that misclassify an emergency call as nonemergency, or that fail to escalate a caller whose situation deteriorates during an automated interaction, could produce outcomes worse than the staffing shortages the technology was designed to address. The 911 community is inherently risk-averse, and for good reason: errors in emergency response cost lives. NENA’s ongoing discussions about AI integration have emphasized that AI must augment human dispatchers rather than replace them, and that human override capability must be preserved at every stage of the call-handling process.
Data privacy and cybersecurity add further complexity. AI systems that process 911 calls handle sensitive personal information in real time, including caller locations, medical conditions, and descriptions of crimes in progress. Those data streams must be encrypted, access-controlled, and compliant with both NENA standards and applicable state privacy laws. As dispatch centers transition to cloud-based architectures, the attack surface expands, requiring cybersecurity investments that many jurisdictions with constrained budgets may struggle to fund.
The Regulatory Landscape and the Path Forward
NENA hosted a four-hour AI workshop at its 2024 national conference, the organization’s most substantial engagement with the technology to date, and followed up with expanded AI programming at NENA 2025. The association’s chief technology officer, Brandon Abley, led sessions at IWCE 2026 on “Artificial Intelligence Integration in 911 Centers,” signaling that the standards body is moving from exploratory discussion to practical implementation guidance. NENA has also convened an AI Critical Issues Forum to develop frameworks for responsible deployment.
At the federal level, OSHA’s updated National Emphasis Program, effective April 2026 and running through 2031, directs proactive inspections across 55 high-risk industries, a framework that intersects with AI-driven dispatch when heat emergencies, industrial accidents, or natural disasters trigger predictive resource deployment. The Strengthening Mobility and Revolutionizing Transportation (SMART) grant program authorized $100 million annually from 2022 to 2026 for technology projects, including proposals to integrate AI with roadway camera systems to alert 911 centers when crashes occur.
The trajectory points toward a dispatch model where AI handles the first layer of every interaction, whether that means answering a nonemergency call, transcribing a frantic 911 report in real time, translating a caller’s Mandarin into English, or predicting that a Thursday afternoon thunderstorm will generate a 40% spike in calls to a specific district. The human dispatcher remains at the center, but the tools surrounding that dispatcher are becoming substantially more capable. For the roughly 6,000 public safety answering points operating across the United States, the question is no longer whether AI will reshape emergency dispatch but how quickly jurisdictions can deploy it while managing the ethical, financial, and operational risks that come with the transition.
FAQs
How is AI being used in 911 dispatch centers right now?
AI systems currently handle nonemergency call triage, real-time speech transcription, live audio translation for non-English speakers, call prioritization, and predictive analytics that forecast call volume surges. Centers in Arlington County, Virginia; Jefferson County, Colorado; and Charleston County, South Carolina, are among those with active deployments.
Does AI replace human 911 dispatchers?
No. AI tools are designed to handle nonemergency calls and repetitive tasks, freeing human dispatchers to focus on life-threatening emergencies that require empathy, contextual judgment, and real-time decision-making. NENA and the broader public safety community emphasize that AI augments rather than replaces trained telecommunicators.
How severe is the 911 staffing crisis?
Between 2019 and 2022, one in four 911 center positions remained vacant. NENA reported in 2024 that 75% of emergency communications centers lack the budget to expand staffing, and 82% are struggling to fill existing vacancies. AI-driven nonemergency call handling has been adopted in part to offset these shortages.
What are the risks of using AI in emergency dispatch?
Primary concerns include algorithmic bias from historical training data, the risk of misclassifying emergency calls as nonemergency, data privacy and cybersecurity vulnerabilities in cloud-based systems, and the need for consistent human oversight to ensure AI does not make life-or-death decisions without trained operator review.
