EXPOSED
The job is entirely headset-and-screen: answering 911 lines, running scripted EMD/EFD protocols, keying incident data into CAD, and broadcasting on radio — and AI already handles non-emergency call triage, transcription, address geolocation, and automatic translation in a growing number of PSAPs. What resists is the live judgment call: deciding what's actually happening when a caller is screaming, incoherent, or lying, prioritizing units across simultaneous incidents, and keeping a person alive on the line until help arrives. Public-sector employment, union contracts, and extreme institutional risk aversion around life-safety failure slow adoption far more than the technology does.
Mixed — a routine tier and a judgment tier. Half your shift is work a machine can already do — ANI/ALI address capture, typing the caller's complaint into CAD, repeating a plate over the radio, transferring the pole-down call to the utility — but the other half is hearing a caller say 'he's fine' while breathing wrong and deciding it's an overdose anyway, which is why this sits at 10 rather than in the automatable band.
Fully desk- and screen-based. You work seated at a console for 12 hours with a headset, three to five monitors, and a keyboard; the only physical demands are shift length, bladder control, and not leaving the position uncovered — nothing about the job requires you to be anywhere but that chair.
Certification preferred, not legally required. Most states require you to hold ENP/EMD certification and pass continuing education, and Priority Dispatch protocol compliance is audited on your calls — but the licence is a certificate, not a professional practice licence, and when a dispatch goes wrong it's the agency and the responding officers who get sued under state tort claims acts, not you personally, which caps this at 8.
The human relationship is the product. The caller has never met you and won't remember your name, but for the four minutes you're talking a mother through compressions on her infant, your voice is the entire intervention — and on the radio side, the units on your channel know your cadence and trust your traffic in a way they won't extend to an unfamiliar voice mid-pursuit.
Exists to be accountable for ambiguous calls. You decide, in seconds and with incomplete information, whether a call is a welfare check or an active shooter, whether to hold the last available unit or send it, and whether to override the protocol card because what the caller is describing doesn't match any of the answers — and those calls get reviewed against outcomes, not against process.
Has AI actually changed your work?