■ CRITICAL RISK ■ Administrative
Yes — the job's entire product is answers to routine questions, which is the single thing modern AI does most cheaply. What remains is the front-desk human presence some institutions still want, and want less of every budget cycle.
“Google is the world's information clerk. You're the world's redundancy.”
Our AI replacement risk score — how we score jobs
Information clerk is a catch-all title covering the person at the counter, desk, or phone who tells you where to go and what you need: the municipal office clerk explaining which forms accompany a permit application, the hospital information desk directing visitors to the right wing, the university registrar's window fielding enrollment questions, the records clerk retrieving files on request. The day is a stream of lookups — schedules, procedures, eligibility rules, locations — delivered with patience to a public that didn't read the website.
Every channel of that stream is being automated in parallel. Websites and search absorbed the first wave of questions decades ago. Chatbots and AI assistants on institutional sites now handle the conversational follow-ups — 'which form do I need if the property is jointly owned?' — that used to defeat static FAQ pages and require a human. Interactive kiosks and digital wayfinding cover the walk-up directions traffic. Records retrieval became self-service portals. Government agencies, perpetually understaffed and budget-squeezed, are among the most eager adopters, because every question a bot answers is a queue that doesn't form at a counter.
The resistant slice is the public itself. Every information desk serves people who can't navigate digital systems — elderly visitors, non-native speakers, people in distress at a hospital, applicants whose situation genuinely doesn't fit the form. Institutions with a service mandate keep a human at the desk for them, and the soft work of calming an anxious visitor or spotting someone who needs more help than they asked for doesn't automate. But that's a justification for one staffed desk, not a department of clerks. Oxford's early automation-probability research flagged this cluster of routine information roles from the start, and the intervening decade has mostly proven it right. Hence our 94.
Automatability: our editorial assessment of current and near-term AI capability
The squeeze is active now. Institutional chatbots, self-service portals, and kiosks are standard procurement items, and public agencies backfill fewer desk positions each budget cycle. Through the late 2020s expect steady attrition rather than dramatic layoffs — retirements not replaced, three desks consolidated into one. The remaining posts will be explicitly service-of-last-resort roles for people digital systems fail, which is real work but a fraction of today's headcount.
Steadily, yes — mostly through attrition rather than layoffs. As chatbots, portals, and kiosks absorb routine questions, agencies and institutions simply don't refill desk positions when people retire. The occupation was flagged as highly automatable by early academic work on job automation, and the decade since has played out accordingly. Openings still appear, but the long-term trend is consolidation to skeleton staffing.
Serving the people the technology fails. Every public counter sees visitors who can't manage apps — elderly, disabled, non-English-speaking, or simply overwhelmed people, often dealing with stressful situations like a hospitalized relative or a denied application. Reading that distress, adjusting your approach, and taking responsibility for getting them sorted is human work. Institutions keep desks open for exactly this; they just need fewer of them.
As a way into an organization, it can be worthwhile — government and university clerk roles often lead to better-protected positions in casework, licensing, or administration. As a long-term destination, no: the question-answering core of the job is being automated on every channel at once. Take it, learn the institution's systems, and move toward roles requiring discretion or specialized knowledge within two or three years.
Faster than government's reputation suggests. Budget pressure makes automation attractive precisely where staffing is thinnest, and agencies at every level are deploying AI assistants for permits, benefits, and records questions right now. The rollout is uneven — some counters will feel unchanged for years — but each successful deployment becomes the template for the next. The human desk persists as the escalation path, staffed accordingly thin.