What the work actually involves
You are not answering real calls. You are supplying the judgment that a model doesn't have. A typical task set asks you to read a simulated caller transcript — a domestic disturbance, a possible overdose, a caller who won't give an address — and assess whether the model's response gathered the right information in the right order, assigned a defensible priority, and routed to the correct resource. Other tasks run the other direction: you author the scenario, including the messy details (background noise, a caller minimizing, conflicting addresses), and write the reference answer plus the reasoning a reviewer can audit.
Expect to work in written form almost entirely. Your notes matter as much as your ratings — "wrong" is not useful feedback, but "the model dispatched before confirming caller location, and location-first is the standard because the call can drop" is. Protocol citation is welcome where it applies (APCO/NENA guidance, EMD card logic, your agency's SOPs), as is flagging where legitimate practice varies by jurisdiction rather than pretending one answer is universal.
What the screen is looking for
micro1 uses an AI interviewer for the first pass. It probes for verifiable specifics — where you dispatched, what CAD system, what console, how many years, which certifications — and then follows up inside your answers. Candidates who speak in general terms about "staying calm under pressure" tend not to advance; candidates who can walk a call from ring to clear, naming the decision points, do. Prior AI experience is genuinely not required, but the ability to explain your reasoning in writing is.
Logistics
- Remote, contractor, no shift coverage or on-call
- Largely asynchronous with occasional calibration sessions
- Flexible hours; project volume varies, often 10–25 hrs/week
- Observed pay band $30–55/hr, set by project and depth of experience — not guaranteed