مقيم ممارسات القطاع العام

The role

We're hiring working public sector professionals to audit AI work product in their own field, not to produce it.

You'll see an AI agent's complete attempt at a realistic task from your job: the task as written, what the agent did, the files it produced, and the scoresheet an AI grader filled in afterwards. You bring the judgment of someone who does this work for a living, and answer four questions:

1. Is this a real task? Would it come up in this job, described this way? Realistic, borderline, or contrived.

2. Is the domain content right? Are the standards, figures, regulations and constraints accurate and consistent, or is something subtly wrong in a way only a practitioner would catch?

3. Does the grading hold up? The grader answered eleven yes/no questions about the finished work: correct format, fabricated figures, requirements met, terminology, and so on. You go through them with the files open and say where you disagree, and why.

4. Is the score defensible? Band the work yourself. Could a practitioner use it? Accept it? Then check the reported score against your band.

About an hour per item. You never perform the task or redo the agent's work.

Who we're looking for

Five roles. Apply to the one that's actually yours:

• Compliance Officers. 4+ years in a regulatory, licensing or program-compliance role, writing findings someone had to act on.

• Administrative Services Managers. 5+ years running facilities, records, procurement or general services for a public agency or institution.

• Child, Family and School Social Workers. Licensed where your state requires it, 4+ years carrying your own caseload.

• Recreation Workers. 4+ years running programs for a parks department, community center, camp or school district, including the staffing and the safety plans.

• First-Line Supervisors of Police and Detectives. 4+ years as a sworn supervisor at sergeant rank or above.

We want people who know the field cold: anyone who has written a finding, a case plan, an incident review or a corrective action that someone else was accountable to. Less of a fit are policy analysts who researched programs without ever administering one, and private-sector equivalents, since these tasks turn on public procurement, mandated reporting and public-records rules.

Also required: near-native business English, because you write a short reason for every judgment, and enough comfort with a spreadsheet or PDF to check whether the numbers in it reconcile.