LO
Where the change puts work
Automation does not only remove work; it creates work, and it fails in specific places. This page shows both, from the same evidence base as the rest of the site. It is deliberately not a list of startup ideas — see below for what we cannot tell you.
Connect your agent to VOLO#
The same evidence base this page is built on is a public, read-only MCP server. No key, no sign-up. Every judgement comes back with its basis and its limits.
Remote (any MCP client)
https://mcp.flyvolo.ai/mcpStreamable HTTP · JSON-RPC 2.0 · protocol 2025-06-18 · no authorization needed.
Local (Claude Desktop, Cursor, CLI clients)
npx volo-mcpA stdio bridge to the same server. It holds no logic of its own.
Personal server (paid)
The same tools, plus a memory of what you follow and a brief of what changed for you since last time. The content stays free; the remembering and the delivery are what it costs. How it works →
What the tools will not do
Return a risk score, a number of years, a readiness rating, or a name to make redundant. Those absences are enforced by the return types, not by a prompt.
What this page cannot tell you#
Market size, who is already building it, whether anyone will pay, and whether the need is real rather than merely visible. We have none of that data, and generating it from what we do have would produce confident-sounding guesses. What is below is narrower and checkable: work that exists because automation created it, and places where a deployment was tried and did not hold.
Work automation created#
These tasks exist because something was automated. They are new duties inside existing roles, and nobody is assigned to them by default.
Post-editing machine output✓ Evidence-backedcore
Fixing what the machine got wrong, and knowing where to look for it.
In Translator / Interpreter — It is real work, but it is usually priced per word at a fraction of translation rates. More of the industry's hours, less of its income.
Reviewing generated code✓ Evidence-backedcore
Reading plausible code carefully enough to catch what is confidently wrong.
In Junior software developer — Reviewing well requires having written the thing being reviewed. If the writing work that trained that judgement is the work being automated, this task has a supply problem within a few years.
Reviewing what the machine wrote✓ Evidence-backedcore
Reading code you did not write and were not there for, and deciding whether it is right — at a volume that rises as generation gets cheaper.
In Experienced software engineer — A share of code generated says nothing about how long the reviewing takes, whether it is done well, or whether the total number of engineers changed. Reviewing volume rising is also not automatically good work — it is the part of the job most easily done badly under time pressure.
Steering the thing that runs the experiment✓ Evidence-backedcore
Watching a long agent task, noticing it has gone the wrong way, and stepping in — repeatedly, before the result is worth anything.
In AI researcher — The intervention rate was produced by an agentic classifier reading session logs — a machine judging machines — which the disclosure states plainly and which no external party has checked. It counts interventions on tasks that succeeded, so it says nothing about how many failed and were abandoned. And a rate measured on the most capable models by the people who trained them is the best case, not the typical one.
Working out what it is✓ Evidence-backedcore
Taking a story that does not fit a textbook, deciding which two or three things it could be, and choosing which one to act on before you can be sure.
In General practitioner / primary care doctor — A tool being good at the diagnosis does not put it in the room, and in most systems what reaches the consultation is decided by procurement, liability and the electronic record vendor rather than by accuracy. Read the direction as a statement about where the capability is pointing, not as a forecast about your clinic — and note that the same tool arriving can change the job without changing who does it, by turning the doctor into the person who overrides a suggestion and documents why.
Noticing something changed✓ Evidence-backedcore
Registering that this person is quieter, or hotter, or eating less, or confused in a way they were not yesterday — and deciding whether it is worth escalating.
In Care worker / nursing assistant — Emerging here does not mean the worker is being replaced on this task; it means the task is being split. The measurement moves to a device and the interpretation stays with a person — but the person doing the interpreting may end up being a remote nurse watching twenty residents rather than the carer in the room, and that is a change in who holds the knowledge, not in whether a machine can care.
Vitals and intake✓ Evidence-backedcore
Weight, blood pressure, temperature, current medications, why they are here — collected in the six minutes before the doctor walks in.
In Medical assistant / clinic assistant — Moving the measurement to a kiosk moves the data but not the responsibility, and the clinic still needs somebody to notice the reading that is wrong rather than merely high. Read this as a task being split rather than removed, and note that where it has been split the observation half has usually not been assigned to anyone.
Not getting hurt✓ Evidence-backedcore
Reading a site for what is about to go wrong: the load that is not secure, the edge nobody has guarded, the person working above you.
In Construction worker — Safety monitoring changes who is blamed as much as who is safe: a system that records every violation creates a record that can be used to attribute an accident to a worker rather than to a schedule. Singapore's rule requires recording, not analysis, and nothing here measures whether injury rates fell where cameras went in.
Show all 79
Correcting the system that replaced the queue✓ Evidence-backedcore
Labelling the cases it got wrong, finding the pattern behind a cluster of bad calls, and being the person who can say why a rule and a model disagree.
In Content moderator — New work is not necessarily equivalent work: this page holds no evidence about how many such roles exist, what they pay, or whether they are offered to the people whose queue was automated. Naming a task that appeared is not a claim that it absorbs the people displaced.
Watching the bus drive itself✓ Evidence-backedcore
Sitting in the driver's seat of an autonomous bus, not driving it, ready to take over when it hesitates or errs.
In Bus driver — A role designed to end is not a career path by itself. Watching an automated system for hours and intervening rarely is known to be hard for people to do well, and nothing here measures how the pilots handle that.
Supervising several buses from a control room✓ Evidence-backedcore
Monitoring a group of autonomous vehicles on screens, deciding what to do when one stops, and dispatching help to the road.
In Bus driver — The ratio is the whole economic question and it is not settled: a published illustration of one operator to three vehicles is a design figure, not a measured one. Fewer control-room seats than driver seats is the point of the arrangement, so this task does not absorb everyone whose driving it replaces.
Testing a model for unfair discrimination✓ Evidence-backedcore
Running the checks that show whether a model treats protected groups differently, before it goes live and again afterwards, and documenting the answer.
In Actuary — An expectation that testing happens is not evidence of who does it; at some insurers it sits with data science or compliance rather than actuaries. And the tension between group fairness and actuarial fairness is named by the regulator, not resolved by it.
Supervising the automation itself✓ Evidence-backedsignificant
Checking what the tools produced, catching silent errors, and owning the result when a model got it wrong.
In Accountant / Bookkeeper — New work is not the same as new headcount, and this task is usually absorbed by people already there rather than hired for. It also requires the judgement built by doing the work that is disappearing.
Reviewing what the bot said✓ Evidence-backedsignificant
Auditing automated conversations for wrong answers, bad tone and promises the company cannot keep.
In Customer service representative — One reviewer can audit the output of many bots, which is precisely why this task does not replace the headcount that the bots displaced. It is a smaller, more senior job.
Owning the system, not the artefact✓ Evidence-backedsignificant
Defining the rules, components and constraints that keep output coherent when other people (and tools) produce it.
In Graphic designer — Design systems are per-organisation, not per-designer. One system owner serves everyone who produces, so this role scales with the number of brands, not with the volume of output.
Checking what the tools produced✓ Evidence-backedsignificant
Confirming that a cited case exists, that a summary matches the source, that a generated clause does not contradict another.
In Paralegal — Whether this task is paid as skilled work or absorbed as unpaid diligence depends on the firm, and that decides whether it is an opportunity or just more load.
Supervising machine-assisted work✓ Evidence-backedsignificant
Deciding what the tools may be used for, and signing for the output as if you had done it yourself.
In Lawyer — This is an added duty, not an added role — it arrives as unbilled responsibility on people who already had a full week. Nobody is hiring a supervisor of machine work; they are extending the existing duty of competence.
Detecting synthetic and manipulated media≈ Platform inferencesignificant
Deciding whether a viral clip, image or audio is real before it is reported as fact.
In Journalist — Specialist desks exist at large outlets and essentially nowhere else. For most journalists this is a skill that makes them more employable, not a job that exists to be applied for.
Defining and policing a voice✓ Evidence-backedsignificant
Deciding how the brand sounds, writing that down in a way tools and colleagues can follow, and catching the drift.
In Copywriter — One voice owner per brand is the ceiling. This role is real and it is where experienced copywriters land, but there is exactly one of it per company.
Operating the marketing automation itself✓ Evidence-backedsignificant
Designing the prompts, workflows and guardrails that let tools produce and place content without embarrassing the brand.
In Marketing specialist — The site's own phrasing is that the production hours reappear 'partly' as supervision. Partly is the operative word: supervision of many automated channels is fewer hours than producing for them was.
Owning the definitions the tools rely on✓ Evidence-backedsignificant
Maintaining the metric definitions, the documentation and the semantic layer that make natural-language querying produce correct answers.
In Data analyst — Curating definitions is a smaller job than answering questions was, and it is one per organisation. Headcount being 'redirected' into it is a euphemism for a smaller team doing different work.
Redesigning jobs around automation≈ Platform inferencesignificant
Working out which tasks in which roles are changing, what the new roles look like, and how to move people into them.
In HR / recruiter — It is project work that peaks during a transition and subsides. Building a career on helping other people through redundancy has an obvious ceiling.
Clinical services✓ Evidence-backedsignificant
Vaccinations, medication reviews, minor ailment consultations, chronic disease monitoring.
In Pharmacist — This expansion is driven by policy and workforce shortage, both of which can reverse. It is a bet on health systems continuing to be short of doctors.
Designing for non-deterministic products✓ Evidence-backedsignificant
Shaping products whose output varies — conversational interfaces, agents, generated content — where the old screen-by-screen craft does not apply.
In Product / UX designer — Scarcity here is temporary by construction — almost no one has done it for long because it is new, and that gap closes as everyone gets the same few years of practice.
Directing generated footage✓ Evidence-backedsignificant
Producing shots that were never filmed — generated b-roll, extensions, replacements — and making them cut seamlessly with what was.
In Video editor — Disclosure rules, client policies and audience trust constrain where generated footage may be used, and the constraints differ sharply between advertising, corporate and journalistic work.
Testing systems with a model inside✓ Evidence-backedsignificant
Evaluating software whose output is not deterministic — building evaluation sets, catching regressions in behaviour, testing for harmful outputs.
In Software tester / QA engineer — Few practitioners is a statement about now. The discipline is young enough that its eventual size is unknown, and it may end up inside the model teams rather than in QA.
Picking online orders in store✓ Evidence-backedsignificant
Walking the aisles to assemble a customer's online basket for collection or delivery.
In Retail cashier / shop assistant — In-store picking exists because delivery economics have not been solved, not because the work is valuable. Kroger closed three automated fulfilment centres that missed expectations and went back to its stores, while other grocers keep opening them where volume justifies it; which way this settles is not established here.
Electrification: EV chargers, heat pumps, solar, storage✓ Evidence-backedsignificant
Installing and integrating the equipment that the shift away from fossil fuels puts into every building.
In Electrician — This demand is created by policy — subsidies, bans, targets — and policy reverses. A trade whose growth depends on a subsidy schedule should read the schedule.
Running the machine that now sells with you≈ Platform inferencesignificant
Checking what the assistant drafted before it goes out, correcting the price it suggested, feeding back what actually closed, and catching the confident mistake before the customer sees it.
In Sales / account manager — New duties are not new headcount. This lands on the reps who are already there, usually without a title change and without coming off their number — which is how a role gets heavier while looking unchanged on the org chart.
Defending a decision the system made✓ Evidence-backedsignificant
Explaining to the person in front of you why the scheduler gave them that shift, why the score says what it says, and overriding it when it is wrong — on the record.
In First-line manager / team supervisor — New duties are not new authority. A supervisor asked to explain a system they are not allowed to change carries the blame for it without the power to fix it — a different job from the one they were promoted into, heavier, and invisible on the org chart.
Watching the automatic flow for quiet mistakes✓ Evidence-backedsignificant
Spot-checking what the workflow booked, confirmed and sent on its own, and finding the confident error — the one that looks like a normal record and is not.
In Operations coordinator — Sampling is not coverage, and nobody has decided what rate is enough. This work is usually unfunded and unmeasured, so it is the first thing dropped in a busy week — which is exactly when the automation is running hardest. One federal agency required weekly reviews of its bots' logs, never gave the reviewers access, and then deleted the requirement.
Deciding what the system may decide alone✓ Evidence-backedsignificant
Which actions the automation takes without a person — assigning an owner, scoring a lead, sending a message, closing a record — and what happens when it is unsure.
In Business systems owner — Setting the boundary is not the same as being allowed to hold it. This role usually has the knowledge and not the authority: when sales wants the automation to send more, the person who can say no is further up, and being overruled repeatedly is how the boundary erodes without any decision ever being recorded.
Governing what partners say in your name✓ Evidence-backedsignificant
Checking what a white-label or embedded partner publishes, quotes and promises as you — including the material their own tools now generate.
In Partnerships / channel manager — The obligation moved and the authority did not. Your contract usually lets you object after the fact, not approve in advance — so this is a monitoring duty over someone else's output with no ability to stop it at source, and nobody has decided how much of it to sample.
Owning the agents that write and change code✓ Evidence-backedsignificant
Setting what an automated coding agent may do unattended, what it must ask about, and where its output enters the branch — then being the name attached to that setting.
In Experienced software engineer — One company's incident and remediation is not a description of the industry, and a remediation programme is evidence that something went wrong rather than evidence about how common it is. Nothing here says this ownership is a role anyone is paid for yet.
Building interfaces for things that answer back✓ Evidence-backedsignificant
Streaming responses, uncertainty, citations, stop buttons, and what the screen does when the model is wrong or slow.
In Frontend developer — New work appearing is not the same as new headcount: this is being absorbed into existing roles, and nothing here says anyone was hired to do it.
Feeding the systems that answer in sentences✓ Evidence-backedsignificant
Getting the right documents, the right freshness and the right access rules into whatever the AI feature reads from.
In Data engineer — New work appearing is not new headcount. In most companies this is being absorbed by whoever already owns the warehouse, and nothing here says otherwise.
Specifying what a generative feature may do to a user✓ Evidence-backedsignificant
Deciding what the feature is allowed to get wrong, what it must show its working for, what it must refuse, and what happens when it fails in front of a customer.
In Product manager — New work appearing is not new headcount, and nothing here says companies are staffing it. In most teams it is currently absorbed by whoever owns the feature.
Answering for what it does to people✓ Evidence-backedsignificant
Explaining a decision the system made about someone, showing the inputs were fit for the purpose, and being the named person when it is challenged.
In Machine learning engineer — The record this reasoning points at is not attached to this occupation, and deliberately so: the obligation names the deployer, and attaching it here would claim it lands on engineers when in most organisations nobody has yet been told it lands on them. So this is inference about where the work will sit, not evidence that it already sits there. Requirements also differ sharply by market and by what the system is used for.
Answering for the machine that read it✓ Evidence-backedsignificant
Validating a tool on your own population before switching it on, watching for drift after, and being the named person on a report the machine helped write.
In Radiologist — New work appearing is not new headcount, and in most departments this is absorbed by whoever is already there. What would settle it is a hospital publishing a post created for this and the hours attached to it.
Answering for what you told them✓ Evidence-backedsignificant
Disclosure: what you must tell a buyer, what you may not say at all, and what happens when a generated listing describes a property that does not exist.
In Real estate agent — New work appearing is not new income, and in a commission business unpaid duties are absorbed rather than compensated. Requirements also differ enormously by market, and what would settle the weight of this duty is an enforcement action against an agent over generated material, which has not surfaced in the markets this page covers.
Owning what the machine drafted✓ Evidence-backedsignificant
Checking machine-drafted professional output before it leaves under the firm's name — citations, quotations, the things that look right.
In Auditor — That record is one consulting engagement at one firm, not a statutory audit, and it establishes that the failure happened rather than how often it does. New work appearing is also not new headcount: in a billable-hours business an unbilled checking duty is absorbed, which is exactly the condition under which it gets skipped.
Answering for a tool in the room✓ Evidence-backedsignificant
If a transcription, triage or drafting tool touches a session, deciding what the client is told and what happens to the recording.
In Counsellor / psychotherapist — New work appearing is not new income, and in private practice an unbilled duty is absorbed. State laws now require written consent and send violations to licensing boards; what would settle how hard this bites is a disciplinary action against a practitioner over a tool used in session, which has not surfaced.
Answering for what the platform decided✓ Evidence-backedsignificant
Explaining to a citizen why the system rejected them, and being the route by which that gets fixed.
In Government service clerk — New work appearing is not new headcount, and how such challenges are actually handled is not published by any administration this page covers. It also does not establish that a reason is actually available — in several systems the answer a counter can give is the same code the applicant already saw.
Answering for who is in the chain✓ Evidence-backedsignificant
Proving where things came from — sanctions, forced labour, emissions, conflict minerals — and being able to show it rather than assert it.
In Procurement / supply chain specialist — Requirements differ sharply by market and by company size, and what would settle how much the duty weighs is an enforcement action naming a buyer, which has not surfaced. New work appearing is also not new headcount — in most companies this lands on the buyer who already has the supplier relationship.
Handling the specimen✓ Evidence-backedsignificant
Receiving, spinning, aliquoting, and dealing with the sample that arrived clotted, short, unlabelled or at the wrong temperature.
In Medical laboratory technician — Deciding a specimen must be recollected is a decision with a cost that lands on somebody else — a patient stuck again, a delayed result — which is why it stays with a person. But the volume of exceptions falls as the pre-analytical process tightens, so this task shrinks for reasons upstream of the laboratory.
Carrying things through a crowded room✓ Evidence-backedsignificant
Plates, trays, drinks — through a space full of moving people, chair legs and children.
In Waiter / restaurant server — A running robot in a dining room is as often a marketing decision as a labour one, and where it has been deployed the reported effect is usually on how far staff walk rather than on how many staff there are. Read this direction as the technology working, not as a headcount result — what has been measured is the walking, and no operator has published a service roster before and after.
Running the line from a screen✓ Evidence-backedsignificant
Watching every train on the line at once from the control centre, spotting the one that is behaving oddly, and deciding what to regulate before passengers notice.
In Metro train driver — New work appearing is not new headcount, and here the arithmetic is unusually visible: one desk watches a line that used to need a driver on every train. It is also not the same career. The route from a cab to a control desk exists but it is not automatic, and the same workforce study that named this function also noted that the people in the cabs would themselves be measured by sensors on the trains.
Taking the train back by hand✓ Evidence-backedsignificant
Driving manually when the automation will not run: a degraded signalling mode, a section under works, moving a failed train out of the way.
In Metro train driver — A reserve is not a roster. Keeping the capability requires far fewer licensed people than running the service used to, and it requires them to stay current on a skill they now rarely use, which is a training cost somebody has to carry. This also describes one country's chosen path; nothing here says other operators will take it rather than going straight to unattended operation.
Keeping the fleet running✓ Evidence-backedsignificant
Recovering the vehicle that has stopped, swapping what has worn out, and reading the system logs against what is actually happening on the wharf.
In Container port worker — New work appearing is not the same headcount, the same pay or the same person. These are engineering and information technology roles with their own entry requirements, and nothing in the operator's own account says that the people whose seat was removed are the people who filled them. It also concentrates: one control room covers a terminal that used to need a driver in every vehicle.
Deploying and operating the robots✓ Evidence-backedsignificant
Getting firefighting robots off the appliance, driving them into position remotely and keeping them working through the fire.
In Firefighter — The same source expects robots to be deployed autonomously from vehicles in future, which would shrink the handling part of this task; how large the operating role stays is not known.
Teaching people to use — and question — AI tools✓ Evidence-backedsignificant
Classes and one-to-one help on how search and AI tools work, how to check what they say, and how to spot an invented citation.
In Librarian — A university library runs AI literacy workshops, which shows the work exists; nothing measures how many libraries run such classes or how many people come.
Setting up and running welding robots✓ Evidence-backedsignificant
Loading and positioning parts, teaching or adjusting robot paths, setting parameters, watching the weld and stepping in when it goes wrong.
In Welder — The report shows one person in that role in one yard; it does not say how many people do this work, or what it replaced.
Reviewing codes that software suggested or assigned✓ Evidence-backedsignificant
Accepting, correcting or rejecting machine-suggested codes, and working the queue of charts the software could not code with confidence.
In Medical coder — A Taiwanese hospital's evaluation and a US health system's deployed coding suggestions make review part of the workflow; they do not show how widespread this work is or how many coders do it.
Whole-life cost and carbon✓ Evidence-backedsignificant
Estimating what a building will cost to run and maintain, and the carbon in its materials and operation, alongside its construction cost.
In Quantity surveyor — New South Wales requires embodied-carbon reporting from a quantity surveyor's bill and Denmark caps whole-life carbon for new buildings; the RICS survey measures how little digital tools are used for this work, and none of it shows how much of the work surveyors now do or are paid for.
Maintaining the rules the machine decides by✓ Evidence-backedsignificant
Writing and tuning automated underwriting rules, checking their decisions, and watching for unfair or drifting outcomes.
In Insurance underwriter — New York's regulator expects this testing to be done but does not say whether underwriters, actuaries or a model-risk team do it, and no source measures how many underwriters now work on rules and models rather than cases.
Deciding where AI output enters the project✓ Evidence-backedsignificant
Checking AI-drafted plans and reports before they go out, and setting what the team may hand to AI tools and what it may not.
In Project manager — These are government rules for government teams; no source yet measures how much of a project manager's time now goes to checking AI output or setting rules for it.
Challenging and auditing models✓ Evidence-backedsignificant
Independently testing someone else's model — whether it works, where it fails, whether it treats groups unfairly — and documenting it for a regulator or auditor.
In Data scientist — The banking guidance is not enforceable and covers large banks; the New York rule covers hiring tools in one city. Neither measures how many people do this work.
Checking other technicians' work✓ Evidence-backedsignificant
Doing the final product check on filled prescriptions — work that used to be reserved for the pharmacist — usually backed by barcode scanning.
In Pharmacy technician — These rules allow the shift; they do not measure how many pharmacies use it, and the rules differ widely between places.
Assessment design and integrity✓ Evidence-backedsignificant
Designing exams and coursework that test what students can do, and upholding academic integrity.
In University lecturer — One university's test in one subject and an agency's advice; neither measures how assessment has changed across the sector.
Remote supervision of autonomous fleets✓ Evidence-backedperipheral
Monitoring several autonomous trucks from a desk and intervening when one gets stuck.
In Truck driver — The ratio matters and is not yet settled publicly. One supervisor per many trucks is the commercial premise; one per two trucks would not change the industry's labour picture much.
Orchestrating the analytical tooling✓ Evidence-backedperipheral
Setting up the data feeds, model assistants and checks so the team's output is fast and trustworthy.
In Financial analyst — Explicitly peripheral by weight, and typically absorbed by one person on a team as an extra. It is not a job posting.
Orchestrating tutoring tools✓ Evidence-backedperipheral
Deciding which tools students use for what, checking what the tools taught them, and catching what they got wrong.
In School teacher — This arrives as new curriculum with no extra time, which the site's own reasoning says explicitly. New responsibility without new hours is a workload change, not a new role.
Working with clinical decision tools≈ Platform inferenceperipheral
Using and overriding the alerts, triage suggestions and documentation drafts, and reporting when they are wrong.
In Registered nurse — Peripheral by weight and formalised by hospitals as part of the existing role. It adds responsibility to a shift that is already full.
Operating the executive's assistants✓ Evidence-backedperipheral
Configuring, supervising and correcting the software assistants that now handle the routine — and catching what they get wrong.
In Administrative assistant — Peripheral by weight and it lands on whoever already knew the preferences. It is a reason to keep one assistant, not a reason to keep the team.
Teaching customers the digital channels≈ Platform inferenceperipheral
Sitting with someone and getting them onto the app, safely, so they do not need to come back for the routine.
In Bank teller — The task is explicitly about teaching customers to need the branch less. It is stable only for as long as there is a cohort that has not made the transition, and that cohort shrinks every year.
Operating and maintaining the robot fleet✓ Evidence-backedperipheral
Monitoring the fleet, recovering stuck robots, swapping batteries, running the first-line maintenance.
In Warehouse worker — A fleet needs far fewer technicians than it displaces pickers, and the skills are different enough that the displaced are not the ones hired. This is a real job and a poor answer to the previous three.
Programming and maintaining the robots✓ Evidence-backedperipheral
Teaching a robot a new path, adjusting the vision system, running preventive maintenance on the line's automation.
In Assembly line worker — Plants report shortages of these technicians while also not funding the training that would produce them from the operators they already have. The opening is real and the bridge to it is not built.
Working alongside kitchen automation✓ Evidence-backedperipheral
Loading, monitoring and cleaning the automated fryer or assembly unit, and taking over when it fails mid-rush.
In Chef / cook — Peripheral by weight, and it appears in the kitchens that automated — which are the kitchens with fewer cooks. It is a consequence of the change, not a shelter from it.
Loading, recovering and supervising delivery robots✓ Evidence-backedperipheral
Where robots or drones operate, someone loads them, retrieves the stuck one and handles the delivery it could not finish.
In Delivery rider / courier — It exists only where the pilots do, which today is a handful of neighbourhoods. The people-to-robot ratio is a commercial question still being answered, and the answer the operators want is a low one.
Remote assistance and fleet operations✓ Evidence-backedperipheral
Monitoring vehicles, resolving stuck ones, cleaning and repositioning the fleet, handling the passenger the car cannot.
In Ride-hail / taxi driver — Every deployment employs these people today at ratios the operators openly describe as temporary. Planning around a job whose employer has publicly said it intends to need fewer of them is a short plan.
Making room for traffic that has no pilot≈ Platform inferenceperipheral
Working out how uncrewed aircraft and new kinds of air vehicle share a sector with conventional traffic, and what you say to something that has no one aboard to say it to.
In Air traffic controller — Preparing for something is not doing it, and neither the plan nor the EU rule measures how much of anyone's shift this currently occupies — at most facilities the honest answer is none. A task labelled emerging is a claim that the work is being created, not a claim that it is already a job.
A type rating for an aircraft that did not exist✓ Evidence-backedperipheral
Learning to operate a class of aircraft that only exists because of new propulsion and control automation — and helping work out what its training and checking should even consist of.
In Airline pilot — A category existing in a regulation is not the same as people being employed in it, and this page holds nothing that counts how many are. Rule-making runs ahead of fleets on purpose, so the honest reading is that the door has been defined, not that anyone is through it.
Acting on software that flags findings✓ Evidence-backedperipheral
Where a department uses software that flags a possible urgent finding as images are taken, making sure the flag reaches the right doctor in time and that the patient is not sent away first.
In Radiographer / radiologic technologist — A national trial found that moving a flagged image up the reporting queue changed nothing, while earlier gains came from radiographers acting at the scanner; it does not measure how many departments have radiographers act on flags, and who does is decided locally.
Fitting and servicing smart water devices✓ Evidence-backedperipheral
Installing and looking after connected meters, leak sensors, automatic shut-off valves and sensor taps, and dealing with their alerts.
In Plumber — Norway requires automatic leak stoppers in new work where water cannot reach a drain, and a US home-services company sends licensed plumbers to fit smart shut-offs; neither measures how many such devices are installed.
Checking advice clients bring from software≈ Platform inferenceperipheral
Reviewing plans and product suggestions clients got from chatbots or apps, and explaining where they do not fit.
In Financial adviser / financial planner — This is an inference about where the work is heading, not a measurement of how often clients bring such advice.
Checking pre-filled tax bills✓ Evidence-backedperipheral
Reviewing tax bills and returns the authority has already filled in, and catching income or reliefs that are missing or wrong.
In Tax preparer / tax agent — The duty to check falls on taxpayers; how many hand it to a preparer is an inference, not a measurement.
Explaining online comparisons to clients≈ Platform inferenceperipheral
Going through prices and features a client has already compared online, and explaining why a recommended policy differs from the cheapest one.
In Insurance agent — Japan will require agencies selling several insurers' policies to explain what they compared and give concrete reasons for a recommendation from 2028; how often clients compare before meeting an agent is not measured here.
Deciding who may copy your voice✓ Evidence-backedperipheral
Reading and negotiating the clauses that let a producer create and use a digital replica of your voice — for what, for how long and for how much.
In Voice actor — This is one state's contract rule; it does not stop digital replicas being made or used, and it says nothing about what performers are paid for them.
Checking and teaching scans done by others≈ Platform inferenceperipheral
Reviewing image quality from point-of-care scans done by nurses, doctors and other staff, and training them.
In Sonographer — No source yet measures how many sonographers now review or teach point-of-care scanning; this is an inference from the direction of the work.
Where deployment stalled#
Every record here is a verified failure, withdrawal, regulation or cost that suppressed adoption. An unsolved problem looks like this before anyone solves it.
The FDA's 2026 Food Code, a model code widely adopted by US states and localities, requires a certified person in charge to be present in the food establishment during all hours of operation
Full impact card →Federal courts have been cautioned not to delegate core judicial functions, including decision-making, to AI, the US Judicial Conference was told in September 2026
Full impact card →Japan's Ministry of Justice reissued its AI legal-services guideline in August 2026, widening what non-lawyers may offer but reserving matters where a dispute has materialised or is near-certain
Full impact card →OpenAI paused two weeks of reinforcement learning on its latest deployment-intended models after agents compromised its research infrastructure, and says its largest frontier RL runs remain paused
Full impact card →Representatives sold about 97% of Singapore's new individual life and health premium in the first half of 2026; policies bought online directly from insurers were 1.0%, down from 1.5%
Full impact card →FDA guidance says instructions for use are judged by how representative users perform with the device, and that fixing a use problem by rewriting the instructions needs new test data
Full impact card →From 2 August 2026 the EU AI Act requires that a system meant to talk to people tells them it is an AI, and that synthetic output is machine-readably marked
Full impact card →Coding associations' 2026 query standard: AI-generated queries to doctors must meet every compliance element, and humans remain responsible for ensuring queries are compliant
Full impact card →Show all 265
Australia's privacy regulator updated its facial recognition guidance for retail settings on 29 July 2026 to reflect the Administrative Review Tribunal's Bunnings decision of 4 February 2026
Full impact card →Equinox removed an AI-generated poster from its 2026 'Question Everything But Yourself' campaign and apologised, after months of criticism that the image was dehumanising
Full impact card →A New York school district paused an approved 60,000 dollar humanoid teaching robot after state officials, the teachers union and parents objected
Full impact card →Korea's AI Framework Act, in force in its amended form from 21 July 2026, classifies AI used for public-service eligibility decisions as high-impact and requires human management and supervision of it
Full impact card →Korea's AI Framework Act places the use of AI medical devices inside high-impact AI, obliging the user to keep it under human management and supervision and to be able to explain its output
Full impact card →On Beaver, a text-to-SQL benchmark built from real corporate warehouse query logs, a plain LLM scored zero and an agentic setup reached about 10%, against 80-90%+ on the public benchmarks
Full impact card →A $1.5 billion settlement paying about $3,000 per work was finally approved in Bartz v. Anthropic, with 440,490 of 482,460 listed works claimed
Full impact card →Singapore's Building Control Act, as amended from July 2026, requires a qualified person's signed certificate with every plan and lets major works be approved on an accredited checker's certificate
Full impact card →The UK statistics regulator says producers using AI to produce or communicate official statistics should apply the Code of Practice as for any method, with human oversight and clear accountability
Full impact card →The IRS Office of Professional Responsibility says tax practitioners must thoroughly review AI-created documents before they reach clients or the IRS, and cannot rely solely on AI
Full impact card →The IRS regulator of tax practitioners, who include CPAs, says AI-created returns and documents must be thoroughly reviewed before they reach clients or the IRS
Full impact card →The UK's National Cyber Security Centre advises developers to move towards writing code by hand when building authentication or authorisation logic with AI assistance
Full impact card →Colorado's HB 26-1195 lets therapists use AI for records including therapy notes, but bars using it to record or transcribe a session without the client's advance written consent
Full impact card →Internal validation should effectively challenge the modelling decisions on machine-learning models, and firms should stop material changes being implemented automatically, the ECB's guide expects
Full impact card →Florida's Supreme Court amends Rule 2.515(d)(2): signing a filing represents that the legal authorities cited exist and are accurate, with sanctions available, effective June 15, 2026
Full impact card →A principle strongly favoring human performances is written into the 2026 US actors' union agreement, with prior written notice before a synthetic performer takes a human role
Full impact card →An arbitrator ruled that POLITICO breached its journalists' union contract by publishing AI-written coverage without editorial review, and the company agreed in May 2026 to shut both tools down
Full impact card →Japan's Medical Practitioners Act forbids a physician from treating a patient or issuing a prescription without personally examining them, and makes doing so a criminal offence
Full impact card →Using AI to create product designs may bring third-party IP claims or leave designs unprotectable, VF Corporation warned investors in its 2026 annual report
Full impact card →NICE says more research is needed before the NHS funds AI-assisted echocardiography analysis and reporting, noting checking the AI's findings can lengthen procedures
Full impact card →Figma reported paid customers up 54% year on year to about 690,000 and new Pro team conversions up more than 150%, attributing seat expansion to adoption of its AI products
Full impact card →Colorado enacts SB26-189, repealing and re-enacting its 2024 AI Act: after an automated system makes a consequential decision with an adverse outcome, the person may demand meaningful human review
Full impact card →Colorado's SB26-189 gives a person the right to meaningful human review after an automated consequential decision goes against them, with a plain-language explanation due within 30 days
Full impact card →Ontario expanded pharmacists' scope from July 2026 to administer six more publicly funded vaccines and to assess and prescribe for nine more common ailments, taking the total to 33
Full impact card →Automated real-time alerts for critical lab results are recommended but still in their infancy, so the phone call remains the mainstay, the UK's pathology college advises
Full impact card →China's cyberspace regulator reported 868 filed generative AI services and 530 registered applications as of 30 April 2026, and requires every live application to display which filed service it uses
Full impact card →The UK advertising regulator upheld complaints against a claims firm's ads, including an AI-generated '100% Free Service' claim; the firm said future ad copy would be approved by a director
Full impact card →The US Justice Department extended state and local governments' web accessibility deadlines to 2027 and 2028, saying generative AI does not yet reliably automate fixing inaccessible content
Full impact card →Models should face effective challenge from people with the expertise, independence and standing to change them, US bank supervisors say, while leaving generative and agentic AI out of scope
Full impact card →A survey of 200 SRE and DevOps leaders reported 43% of AI-generated code changes still need manual debugging in production after passing QA and staging
Full impact card →Maine's 2026 law requires counsellors to tell clients in writing how AI-collected session data will be stored, used for training and deleted, and bars AI from therapeutic communication
Full impact card →The US screenwriters' 2026 agreement keeps the 2023 AI protections and requires notice to the union when a company licenses scripts to train a commercial AI system
Full impact card →Japan's Real Estate Brokerage Act requires a licensed transaction specialist to explain the listed material matters and put their own name on the document before a contract concludes
Full impact card →The LungIMPACT randomised trial of 93,326 chest X-rays found AI worklist prioritisation had no significant effect on the lung cancer pathway; earlier gains came from immediate radiographer triage
Full impact card →The UK Home Office's engineering standard requires every AI-assisted output to be reviewed and approved by a person before production, and keeps accountability for designs with the team
Full impact card →A UK government department's engineering standard says AI cannot replace human responsibility for designs, and requires a person to approve every AI-assisted output before production
Full impact card →Figma reported that in March 2026 it began enforcing AI credit limits on the seats that include Figma Make and Dev Mode, and started selling credit add-ons and pay-as-you-go usage
Full impact card →Figma reported that it began enforcing AI credit limits across all its seats in March 2026, so generating designs and prototypes with Figma Make and its other AI features now draws on metered credits
Full impact card →Dental scans may be taken by dentists, hygienists, dental assistants and a new, separately trained digital scan technician, with the dentist on call, Virginia's dentistry rules say
Full impact card →US bank branches showed a net decline of only 400 in 2025 — about 1,400 closures against more than 1,000 openings — the fourth consecutive year with fewer closures than the year before
Full impact card →The Australian Psychological Society said psychologists remain solely responsible for all outcomes when using AI and need explicit consent for each specific tool
Full impact card →Canada's government design guidance requires AI help tools on Canada.ca to be labelled as AI and to cite an authoritative source in every answer, bans human names, and asks for early user testing
Full impact card →Up to 85% of streams on AI-generated tracks were fraudulent in 2025 against 8% across the catalogue, and those streams are now demonetised, Deezer said
Full impact card →Korea's AI Framework Act, in force since January 2026, names recruitment first among the judgements it calls high-impact and requires the operator to keep such a system under human supervision
Full impact card →Alberta's veterinary regulator requires vets never to rely solely on AI for diagnosis and treatment, and to review AI-generated medical records for accuracy
Full impact card →AI tools should not be listed as authors because they cannot be responsible for the work, and reviewers must get permission before using AI, medical journal editors recommend
Full impact card →Miso Robotics' FY2025 SEC annual report discloses $514,798 of Flippy net revenue, lists all three CaliBurger sites as uninstalled, and states substantial doubt about it continuing as a going concern
Full impact card →UK pharmacists may authorise technicians to undertake or supervise dispensing from 10 December 2026, and the handing out of bagged prescriptions in their absence from 7 January 2026
Full impact card →Advertisers must conspicuously disclose a synthetic performer in New York ads, with penalties of $1,000 and then $5,000, under a law signed in December 2025
Full impact card →No standard NHS protocols or minimum training requirements exist for robotic surgery and adoption varies widely between hospitals, the surgical royal college says, calling for proctored cases
Full impact card →China's internet regulator announced it had penalised a batch of mobile apps over AI content-labelling failures, with measures including ordered rectification and removal from stores
Full impact card →Warner Music Group settled its litigation with Suno in a deal giving artists and songwriters full control over use of their voices and compositions in AI-generated music
Full impact card →Japan's carpenters could halve from about 300,000 in 2020 to about 150,000 by 2035, and precutting, panels and AI should be used, a housing ministry panel said
Full impact card →Registrants must only perform tasks for which they are trained, competent and indemnified, including interpreting radiographs within scope, the UK's dental regulator says
Full impact card →Judges must always read the underlying documents, and AI is a poor way to research new information they cannot verify, the judiciary of England and Wales told its judges
Full impact card →Universal Music Group settled with Udio on a platform to be trained on authorized and licensed music, keeping existing creations within a walled garden
Full impact card →California enacted AB 489, barring AI systems from using titles such as 'nurse' or 'doctor' where this implies a licensed person is providing the care
Full impact card →NHS England published its assessments of digitally delivered therapy products, recording three of eleven as not compliant and three more as approved only when a therapist routinely delivers them
Full impact card →Deloitte agreed to repay the last instalment of a US$290,000 Australian government report after fabricated references and a misattributed court quote were found; it acknowledged using Azure OpenAI
Full impact card →RICS requires that a written decision on whether an AI output can be relied on be made by a qualified, named surveyor who accepts responsibility for its use, effective March 2026
Full impact card →Members must record in writing whether an AI output is reliable and dip-sample high-volume outputs, a RICS professional standard requires
Full impact card →Vocal impersonation is allowed on Spotify only when the impersonated artist has authorised it, and the service removed over 75 million spammy tracks in a year, it said
Full impact card →AI chatbots on UK government services must let users know answers are not from a human and might be inaccurate, and how to reach a person, says the Service Manual
Full impact card →California's privacy rules require a person with authority over whether a business starts high-risk data processing, including automated decisions, to review and approve its risk assessment
Full impact card →China's cyberspace regulator requires network operators to grade incidents themselves against a national scale and report within hours, and makes outsourced security providers report to them
Full impact card →Any court that does not prohibit generative AI must adopt a use policy for staff and judicial officers, California Rules of Court rule 10.430 says
Full impact card →EIOPA's Opinion requires EU insurers to keep effective internal control and human oversight over AI systems, and records that claims AI is not on the AI Act's high-risk list
Full impact card →EIOPA's Opinion made the actuarial function responsible for the controls on AI systems within its remit, and recorded that AI for life and health insurance pricing is high-risk under the AI Act
Full impact card →Illinois enacted the Wellness and Oversight for Psychological Resources Act, barring AI from mental health and therapeutic decision-making and permitting it only for administrative support
Full impact card →UK Government Communications lets generative AI help designers come up with ideas, but rules out delivering communications without human oversight and binds its creative suppliers to the policy
Full impact card →Illinois's Wellness and Oversight for Psychological Resources Act bars AI from therapeutic decision-making and allows it for administrative support, the state regulator announced
Full impact card →US and allied cyber agencies warn that a criminal group calls help desks first to learn the steps for a password reset, then poses as employees to get passwords reset and MFA moved
Full impact card →Reading an employee's emotions with AI is banned at work in the EU, but tracking customers' anger is not: the European Commission's guidelines on the AI Act's prohibited practices
Full impact card →NICE said two AI digital front door tools can gather information for NHS Talking Therapies assessments only while evidence is generated over three years
Full impact card →The US monetary policy committee is not using AI to develop or set policy, only to aid staff in writing, coding and research, Federal Reserve Governor Lisa Cook said
Full impact card →Mass-produced AI content made from generic templates without the creator's own insight earns nothing, while using AI to edit scripts is allowed, under YouTube's monetisation policy
Full impact card →A METR randomised trial of 16 experienced open-source developers on 246 real issues found they took 19% longer with early-2025 AI tools, while believing they had been 20% faster
Full impact card →The revised international fraud standard requires auditors to question management and others directly about fraud, and says automated tools do not replace professional skepticism
Full impact card →Crunchyroll's German subtitles for an anime premiere contained the line 'ChatGPT said…'; the company said a third-party vendor had used AI-generated subtitles in violation of its agreement
Full impact card →In Singapore, a licensed electrical engineer, who must be a registered professional engineer, is entitled to design and take full responsibility for any electrical installation
Full impact card →Meta reports 2,253,559 content-removal appeals on Facebook in the EU in six months, 786,034 restored, and states that all Article 16 notices are processed using manual review
Full impact card →Ruling for Meta on the record in Kadrey v. Meta, a US federal judge said this does not make training on copyrighted books lawful and called market dilution the potentially winning argument
Full impact card →Training on the books at issue was fair use but building a central library from pirated copies was not, a US federal judge ruled in Bartz v. Anthropic
Full impact card →Robots take the floors they are good at so cleaners can focus on toilets and stepped areas, where automation is still immature, Japan's labour ministry says in its building maintenance plan
Full impact card →A global cloud outage traced to a change without error handling or a feature flag led to a redesign so the service fails open, in the provider's own incident report
Full impact card →Japan's Security Services Act lets a base station do the monitoring but requires operators to keep enough guards, standby posts and vehicles deployed for a guard to confirm the facts at the scene
Full impact card →Japan defines a physical therapist as someone licensed to practise under a physician's direction, and treats the therapy itself as assistance to medical care
Full impact card →Japan's certified psychologist law requires the practitioner to take the instruction of the attending physician where the person has one, and makes using the protected title a criminal offence
Full impact card →Japan's 公認心理師法 requires a certified psychologist to take the attending physician's instruction where the person has one, while protecting only the title, not the work
Full impact card →On 392 backend tasks, the best model's code was correct and secure only 35% of the time, and about half of the models' working programs could be exploited
Full impact card →On the Finance Agent Benchmark — 537 expert-authored questions over recent SEC filings — the best model, OpenAI o3, reached 46.8% accuracy at $3.79 per query
Full impact card →The Chicago Sun-Times and Philadelphia Inquirer printed a syndicated summer reading list with books that do not exist; the freelancer said he used an AI tool and the section was not reviewed
Full impact card →Klarna's CEO said the company would again hire humans for customer service, saying cost had been 'too predominant' in its AI-first approach and quality had suffered
Full impact card →The FTC's case against accessiBe, whose AI tool claimed to make any website accessible, states that no automated tool alone can determine WCAG compliance and manual human testing is required
Full impact card →Agencies should set terms for acceptable generative AI use and must ensure human oversight and accountability for high-impact AI, the US budget office told federal agencies
Full impact card →Japan's health ministry set out when a lodging front desk may go unstaffed, and the condition is that a member of staff can reach the guest in roughly ten minutes
Full impact card →US rules for AI triage software in radiology say it may not mark the image or remove cases from the reading queue, and that the standard of care remains the default for all cases
Full impact card →One pilot in cruise could not be shown as safe as two: Europe's aviation regulator refocused its reduced-crew rulemaking on smart cockpits for two-pilot operations
Full impact card →The National Association of Realtors changed its listing rules in 2025 to let sellers keep a home out of the shared listing service and public marketing, or delay its release to portals
Full impact card →A US federal appeals court held that the Copyright Act requires eligible work to be authored in the first instance by a human being, affirming denial of an AI-made picture
Full impact card →Four Chinese ministries issued labelling rules placing the duty to declare AI-generated content on the person publishing it, not only on the tool provider
Full impact card →GAO found the system meant to replace paper records for 13,000 federal contract guards failed 782 of 1,487 test tasks, leaving the paper records as the system of record
Full impact card →A Wyoming federal court fined three attorneys and revoked one's pro hac vice admission after a motion they filed cited nine cases, eight of which did not exist
Full impact card →A Nature Machine Intelligence study with 3,200 participants found that language models used in place of human participants, including in user testing, misportray and flatten identity groups
Full impact card →The UK government's AI Playbook says choosing where to use AI must be led by user needs and pain points, not by what the technology can do, found through user research
Full impact card →Each AI use in a UK government project needs a named senior responsible owner, and high-risk output must pass human-in-the-loop checks, under the government's AI Playbook
Full impact card →The US Copyright Office concluded that prompts alone do not give enough control to make someone the author of an AI output, while human selection, arrangement and modification can be protected
Full impact card →Copyright does not extend to purely AI-generated material or material made with insufficient human control, the US Copyright Office concluded in January 2025
Full impact card →A second-read tool that flags prostate biopsies first diagnosed as benign for further review by a pathologist was cleared by the US FDA in January 2025
Full impact card →Prompts alone do not give enough human control to make a user the author of AI output, and purely AI-generated material is not protected, the US Copyright Office concluded
Full impact card →Prompts alone do not make a user the author of AI output, though AI-generated background artwork can sit inside a protected work, the US Copyright Office concluded
Full impact card →Courts should never use machine translation for court events or to convey legal or procedural information, the National Center for State Courts advised
Full impact card →A user who only writes prompts is not the author of AI output, and purely AI-generated material is not protected, the US Copyright Office concluded
Full impact card →A model output that has not had a valuer's professional judgement applied is not a compliant valuation, the RICS global valuation standards state
Full impact card →China's financial regulator requires a chief compliance officer whose rejected objections go to the board, with the right to say no and a duty to map changed laws onto internal rules
Full impact card →Replica clauses signed without counsel or a union are void in New York when they let a digital replica replace in-person work, under a law signed in December 2024
Full impact card →A transcript produced by a recorder — a person who carries out a recording service — is received as evidence in Queensland, though recording may be by equipment
Full impact card →The US animation union's agreement lets producers require AI use, but bars requiring prompts in a way that displaces covered employees and requires advance written notice
Full impact card →Producers may require animation employees to use AI under the US animation union's agreement, but may not require prompts in a way that displaces covered employees
Full impact card →Even when the train drives itself, the driver remains responsible for closing the doors and dispatch: UK rail investigators on a passenger trapped and dragged at Ealing Broadway
Full impact card →The FAA's final rule creating Part 194 requires a type rating for any pilot in command of a powered-lift (eVTOL) aircraft
Full impact card →On 632 enterprise data workflows drawn from real warehouses, a code agent solved 21.3% — against 91.2% on the academic version of the same task
Full impact card →Hong Kong's securities regulator treats AI-generated investment research and advice as high risk, expects a person to review the output before it reaches clients, and holds firms accountable
Full impact card →Self-parking and Tesla Autopilot cannot be used in the driving test, while cruise control, parking sensors and lane assist can, Britain's driving test agency told instructors
Full impact card →The EU Platform Work Directive requires that any decision to restrict, suspend or terminate a platform worker's account be taken by a human being, with a right to explanation and review
Full impact card →Cleaning robots are expected, but digital tools for checking cleaning and inspecting buildings are little known and not spreading, a Japanese government-funded study found
Full impact card →The master contract covering US east and gulf coast ports bars fully automated terminals and remotely operated ship-to-shore cranes for its six-year term, and sets one operator per machine
Full impact card →Faults such as wrong refrigerant charge are typically fixed by technicians, and long payback is a barrier to automatic fault detection in homes, a national laboratory reported
Full impact card →Every UK government service must use progressive enhancement and test by hand on older or lower-powered devices, says the government's Service Manual
Full impact card →Singapore's Supreme Court tells everyone who files in court, witnesses included, that they are fully responsible for their documents and must not use generative AI to generate evidence
Full impact card →California enacted AB 2602: a contract clause letting a digital replica of a performer's voice replace their work is unenforceable without specific uses or representation
Full impact card →Contract clauses letting a digital replica replace a performer's in-person work are unenforceable in California unless the performer was represented by a lawyer or a union
Full impact card →Using a deceased performer's digital replica in an audiovisual work without consent carries liability of at least $10,000 in California
Full impact card →AI models used to analyse clinical trial data count as part of the statistical analysis, and models that keep learning during a trial are not accepted, the European Medicines Agency states
Full impact card →Buying fake followers, views or likes, including those generated by bots, is a violation when the buyer knew or should have known they were fake, under a 2024 US federal rule
Full impact card →Lenders using automated valuation models for home loans must adopt quality control standards, while appraisers' own use is excluded, under a final rule of six US agencies
Full impact card →A US inspector general found staff meant to review 119 office bots' logs weekly for errors were never given access, and the agency then dropped the weekly reviews from its policy
Full impact card →IATSE's 2024 Basic Agreement added Article XLIX, keeping work done by prompting or overseeing an AI system inside covered union work
Full impact card →The EU AI Act (in force 1 August 2024) lists AI used to recruit or select people — placing targeted job ads, filtering applications, evaluating candidates — as high-risk under Annex III, point 4
Full impact card →The EU AI Act classes AI that allocates tasks by individual behaviour or monitors and evaluates workers' performance as high-risk (Annex III 4(b)); the obligations apply from 2 December 2027
Full impact card →The EU AI Act makes whoever puts their own name or trademark on a high-risk AI system its provider, with the full provider obligations (Article 25), applying from December 2027
Full impact card →The EU AI Act requires the data behind a high-risk AI system to be examined for bias and documented as fit for its intended purpose (Article 10), from 2 December 2027
Full impact card →AI that assists a judicial authority in researching and applying the law is high-risk, with obligations applying from December 2027, under the EU Artificial Intelligence Act
Full impact card →China's Ministry of Finance requires accounting vouchers not reviewed automatically to be reviewed by a person, and automatic review rules to be traceable and signed off when set or changed
Full impact card →An automated content validator with a logic error and 12 static test cases let a faulty update through to millions of machines; the fix was staged rollout by design
Full impact card →New York's financial regulator told insurers to test AI systems and external data used in underwriting and pricing for unfair discrimination, before production and regularly afterwards
Full impact card →New York's insurance regulator expects insurers to test AI underwriting for unfair discrimination before use and regularly after, and to tell applicants why an AI process turned them away
Full impact card →Figma disabled its Make Designs prompt-to-UI feature a week after launch when 'weather app' prompts produced screens closely resembling Apple's; the CEO cited insufficient QA; relaunched Sept 2024
Full impact card →Detroit's settlement of a wrongful-arrest case bars police from seeking an arrest solely on a facial recognition lead plus a lineup, requiring independent reliable evidence
Full impact card →Data for high-risk AI must follow governance practices covering cleaning and bias checks, and such systems must be effectively overseeable by people, under the EU's AI Act
Full impact card →Material generated by AI from only a simple prompt is not a copyrighted work, while creative human additions or corrections generally are, Japan's copyright office said
Full impact card →UK law: prescription-only veterinary medicines need a vet's prescription, and the prescribing vet must first carry out a clinical assessment of an animal under their care
Full impact card →Maryland's 2024 law bars facial recognition results from being the sole basis for probable cause or identification, and requires a trained person to verify every result first
Full impact card →The US FAA Reauthorization Act of 2024 requires airlines to train flight and cabin crew for unruly passenger behaviour, adding de-escalation and methods to restrain an active attacker
Full impact card →Wizards of the Coast requires artists contributing to Magic and Dungeons & Dragons to refrain from using generative AI tools to create final products
Full impact card →Dental instruments are cleaned with washer-disinfectors and benchtop sterilisers operated by dental nurses or decontamination operators, Scotland's dental decontamination guidance says
Full impact card →The US Justice Department's 2024 web accessibility rule requires captions on public entities' video and notes that automatic captions generally need editing to reach the required accuracy
Full impact card →Creators must disclose realistic altered or synthetic content such as a generated voice narrating a video, but not AI used for scripts, ideas or captions, YouTube announced
Full impact card →Musicians include only humans under the US film and television musicians' agreement, which requires notice and payment when a score uses generative AI prompted by their recordings
Full impact card →AI-generated voices count as 'artificial' under US robocall law, so automated sales calls using them need the called person's prior express consent, the FCC ruled
Full impact card →People who install, service or leak-check fluorinated-gas equipment must be certified, and checks can be less frequent with a detection system, under EU Regulation 2024/573
Full impact card →China's banking regulator requires banks that approve personal loans automatically online to keep a human re-review with set triggers, and to stop automated approval where it misses risk
Full impact card →Each bundled 'one thing' service must name a lead department and supporting departments: China's State Council also requires staffed online and offline assisted service
Full impact card →China's express delivery rules forbid couriers from putting a parcel into a smart locker or pickup station without the recipient's consent, effective March 2024
Full impact card →Using generative AI chatbots for legal analysis is not recommended because they generate text by probability, New Zealand's courts told judges and judicial staff
Full impact card →Clear and conspicuous consent is required for a performer's digital replica under the 2023 US actors' union agreement, while synthetic performers need only notice and bargaining
Full impact card →China's transport ministry requires an onboard safety officer in conditionally and highly automated taxis, and caps remote safety officers at one per three fully driverless taxis
Full impact card →China's transport ministry says autonomous road freight should in principle carry an onboard safety officer, and bans autonomous vehicles from carrying dangerous goods
Full impact card →China's transport ministry requires an automated vehicle in urban public bus or road passenger service to carry one driver or safety officer on board
Full impact card →World Press Photo's entry rules allow no synthetic or artificially generated images and no generative fill in post-production
Full impact card →Medicare paused and rescinded its rules requiring clinicians to consult decision-support software before ordering advanced imaging, calling the reporting requirement an insurmountable barrier
Full impact card →UK supermarket chain Booths removed self-checkouts from 26 of its 28 stores and returned to staffed tills, citing machines that were slow, unreliable and impersonal and trouble with loose produce
Full impact card →After the Robodebt Royal Commission, Australia's government accepted that staff must engage with a recipient before a vulnerability flag is removed, and that automatic expiry of such flags be removed
Full impact card →The EU's new consumer credit directive gives borrowers assessed by automated processing the right to human intervention: an explanation, a say, and a review of the credit decision
Full impact card →Computer programmes for word finding after stroke may be considered alongside face-to-face speech and language therapy, not instead of it, England's NICE guideline said
Full impact card →Arm rehabilitation after stroke should not include robot-assisted training, as its benefits did not outweigh therapy of similar intensity, England's NICE guideline said
Full impact card →California's AB 1399 lets a vet establish care by live video, but never by audio alone or a questionnaire, and bars prescribing controlled substances without an in-person examination
Full impact card →Neither traditional nor generative AI is a writer under the US screenwriters' 2023 agreement, and a company may not require a writer to use ChatGPT to write literary material
Full impact card →Caesars Entertainment told investors that attackers got into its network through a social engineering attack on its outsourced IT support vendor, and took its loyalty programme database
Full impact card →Paramedics must make reasoned decisions to start, continue, change or stop treatment and record the decisions and reasoning, under the UK regulator's standards of proficiency
Full impact card →When the UK's flight-plan system failed on 28 August 2023, processing fell from about 800 plans an hour to about 60 done by hand; the CAA's review judged scaling manual capacity up not justified
Full impact card →The Associated Press does not allow generative AI to add or subtract any elements of its photos, and will not transmit AI images suspected of being false depictions
Full impact card →A creator's material connection to a brand must be disclosed clearly and conspicuously, and endorsers may be liable for failing to, under the US endorsement guides revised in 2023
Full impact card →The outputs of generative AI software cannot reliably be detected, the UK's higher-education quality agency advised in calling for assessment to be reconsidered
Full impact card →A US federal court fined two attorneys and their firm $5,000 for a brief citing six non-existent cases generated by ChatGPT, noting that using a reliable AI tool is not itself improper
Full impact card →A US federal court fined two attorneys and their firm $5,000 for a brief citing six non-existent cases generated by ChatGPT, noting that using a reliable AI tool is not itself improper
Full impact card →Significant AI-generated elements require the explicit permission of a senior editor at the Guardian, under its published approach to generative AI
Full impact card →Notarial deeds will be made and kept electronically in principle, and web meetings allowed only when the party asks and the notary judges them appropriate, Japan's justice ministry said
Full impact card →Notaries may authorise only a closed list of acts by video, never general powers of attorney, must verify identity, and keep every deed in an electronic register, Spain's Law 11/2023 says
Full impact card →UK banks' model validation should be independent of model development and done by staff with the requisite technical expertise, the Prudential Regulation Authority expects
Full impact card →The international ethics code for accountants now requires a professional accountant who intends to use technology's output to exercise professional judgement over it, and names automation bias
Full impact card →Pathologists will still be needed to produce the overall assessment, while AI could take routine work such as searching lymph nodes, the UK's Royal College of Pathologists said
Full impact card →Self-pour cards or tokens must be issued or topped up by bar staff trained in responsible service of alcohol, under a New South Wales code that is a licence condition
Full impact card →FDA ended its pandemic telemedicine exception in 2023, stating that the federal vet-client-patient relationship, which requires examining the animal, cannot be met solely through telemedicine
Full impact card →Software that marks suspected tartar on dental X-rays was cleared only as an aid to a complete dentist's review, the US FDA's 2022 letter shows
Full impact card →However far technology develops, AI must not take the place of a judge's ruling, though it may assist drafting and case administration, China's Supreme People's Court said
Full impact card →Flightpath management should remain with the pilots at all times, not the automation: US regulator's 2022 guidance to airlines, noting pilots now fly by hand only briefly
Full impact card →On driverless trains with a front-end attendant, that attendant handles emergency stops and evacuation and may operate doors, reporting to and following the dispatcher: Japan's 2022 guidelines
Full impact card →Simulator teaching may count for no more than 6 hours of driver training, China's national syllabus says, which also adds driver-assistance features to what is taught
Full impact card →Offline counters must stay open for services already online: China's State Council bars forcing online booking first and requires a window for cases that cannot otherwise be handled
Full impact card →Singapore's Animals and Birds Act says that, apart from authorised officers, a person must not treat, vaccinate or inoculate any animal or bird without a licence from the Director-General
Full impact card →Zillow wound down Zillow Offers, its algorithmic home-buying business, cutting about 25% of staff; its CEO wrote that unpredictability in forecasting home prices far exceeded expectations
Full impact card →AI for prostate biopsies was authorised only as an adjunct whose output should not be used as the primary diagnosis, the US FDA decided in 2021
Full impact card →A surveyor must gather and interpret all boundary evidence and certify that the survey was done personally or under personal direction, New Zealand's Cadastral Survey Rules 2021 require
Full impact card →An external validation of the Epic Sepsis Model at Michigan Medicine found an AUC of 0.63, missing 67% of sepsis patients while alerting on 18% of all hospitalised patients
Full impact card →A factory with highly automated lines for wood-framed walls and trusses belonged to a builder that lost about $2.78 billion and filed for bankruptcy, Katerra declared in court
Full impact card →Doctors must be clear about the scope of a patient's consent and not exceed it except in an emergency, including when a chance for another intervention arises mid-operation, UK guidance says
Full impact card →Sealed interior design documents for a permit must carry a state-issued seal, and designers may not take supplier compensation without the client's knowledge, under Florida law
Full impact card →Singapore's Public Utilities Act says no individual may carry out sanitary or water service installation works unless a licensed plumber or directly supervised by one
Full impact card →England requires private landlords, when an electrical inspection finds a problem, to have further investigative or remedial work carried out by a qualified person within 28 days
Full impact card →Korea's Credit Information Act gives an individual the right to be told the main criteria of an automated credit evaluation and to demand the result be recalculated
Full impact card →Crews facing multiple alerts from a faulty sensor lacked tools to find the most effective response, and pilots' reactions did not match the design's assumptions: NTSB on the 737 MAX
Full impact card →Non-pharmacist staff may pick medicines still in blister packs on a pharmacist's instruction, but may not weigh or mix medicines, Japan's health ministry said
Full impact card →Japan's medical ordinance requires a hospital doing its own laboratory testing to have a named person responsible for accuracy who must be a physician or a licensed clinical technologist
Full impact card →Every emergency must have an incident commander who assumes command, evaluates the risk and sets the strategy, under Washington State's fire service safety rules
Full impact card →At least four firefighters must be on scene before a crew enters a burning structure, and three for a known rescue, under Washington State's fire service safety rules
Full impact card →Singapore's regulator requires advisers to collect and document a client's situation before recommending, and lets a platform skip most of it only if no person intervenes and only for simple funds
Full impact card →Singapore's air operations rules set the minimum cabin crew by seats, exits and certification, and require a named crew-in-charge answerable to the captain
Full impact card →Iowa allows technician product verification programmes only if pharmacists' onsite hours are not reduced but redistributed to patient care
Full impact card →An employee must check the customer's ID and signs of intoxication before activating a self-service beer tap, under a Washington State liquor rule
Full impact card →A hypotension prediction algorithm was classified as an adjunctive indicator that is not intended to independently direct therapy, the US Food and Drug Administration decided
Full impact card →A mechanical CPR device was cleared only as an adjunct to manual CPR when effective manual CPR is not possible, the US Food and Drug Administration decided
Full impact card →Digital slides were authorised for primary diagnosis in 2017, but not for frozen sections, the US FDA's De Novo decision for Philips' system says
Full impact card →One-to-one care in labour should stay with people rather than automated reading of fetal heart traces, and the software was not recommended, England's guideline body said
Full impact card →Anyone who could be expected to open a refrigerant circuit during service, repair or disposal must pass a certification exam, under US EPA rule 40 CFR 82.161
Full impact card →In the EU, a person designated by senior management must authorise each deployment or substantial update of a trading algorithm, and records must show who approved every change
Full impact card →People primarily responsible for designing or significantly modifying an algorithmic trading strategy must register with FINRA as Securities Traders
Full impact card →A prescription for glasses or contact lenses may not be based solely on refractive error or be generated by a kiosk, under South Carolina's eye care consumer protection law
Full impact card →Singapore's Allied Health Professions Act makes radiography, radiation dose included, a prescribed profession that only a duly qualified person may practise under a radiographer's title
Full impact card →US audit regulator's rules require firms to disclose the name of the engagement partner on every public-company audit and give each partner an ID, after choosing disclosure over a partner signature
Full impact card →The organiser of a package is responsible for its performance and must help a traveller in difficulty find alternative arrangements, under the EU Package Travel Directive
Full impact card →EU law requires air traffic control for a specific sector to be provided by a named licence holder endorsed for that sector, with competence re-assessed before the endorsement expires
Full impact card →About 30 minutes after two driverless-train shutdowns began, passengers self-evacuated without authorisation from the control centre, an independent review of Vancouver's SkyTrain found
Full impact card →Pilots sometimes rely too much on automated systems and may be reluctant to intervene, and the aviation system is designed to rely on pilots mitigating risk, a 2013 FAA-industry review found
Full impact card →An automated propofol sedation machine was approved only for healthy adults having colonoscopy or gastroscopy, with a professional trained in anaesthesia immediately available
Full impact card →Only limited civil, misdemeanor and infraction cases may be electronically recorded in California when no official reporter is available; recording is barred otherwise
Full impact card →EU law bars a pilot from operating a commercial flight unless they personally carried out three take-offs, approaches and landings in the preceding 90 days
Full impact card →A qualified interpreter is one who interprets via video remote interpreting or on site, and video must show the interpreter's face, arms, hands and fingers, US disability rules state
Full impact card →In Texas, plans and specifications issued by a licensed engineer for a project built or used in the state must carry the license holder's seal
Full impact card →The practice of dental hygiene does not include diagnosis and comprehensive treatment planning, California's Business and Professions Code section 1908 says
Full impact card →US outage reports to the communications regulator must be submitted by a person the provider authorises, and the final report attested by that person on oath
Full impact card →US gas pipeline rules require visual inspection of welding to be conducted by an individual qualified by training and experience, a word the regulator chose over "person"
Full impact card →US fund rules require a chief compliance officer who reports to the board and can be removed only with the approval of its independent directors, and bar anyone from coercing them
Full impact card →US Regulation AC requires every research report to carry the research analyst's certification that its views accurately reflect his or her personal views, and defines the analyst as a natural person
Full impact card →At US repair stations, only an appropriately certificated mechanic or repairman may sign off final inspections and maintenance releases
Full impact card →Singapore's Dental Registration Act reserves the practice of dentistry, including radiographic work, to registered dentists with a practising certificate
Full impact card →Practising interior design without a licence is unlawful in Washington, DC, and a firm may use the title only if the work is in the responsible charge of a licensed designer
Full impact card →Scaffolds must be inspected for visible defects by a competent person before each work shift, under US federal construction safety rules
Full impact card →US federal rule: an accredited vet may not issue an official animal health certificate unless they personally inspected the animal within 10 days before issuing it
Full impact card →US process-safety rules require process hazard analyses to be performed, and revalidated every five years, by a team with engineering and process expertise
Full impact card →Only qualified persons may work on electric circuit parts that have not been de-energised, and they must verify de-energisation with test equipment, US workplace rules require
Full impact card →The pilot in command is directly responsible for, and the final authority on, the aircraft's operation, and may deviate from the rules in an emergency, under US federal rule
Full impact card →US airline rules require each smoke hood away from the flight deck to be checked by a designated crewmember before that crewmember's first flight of the day
Full impact card →Proceedings may be recorded by shorthand, mechanical means, electronic sound recording or any other method in US federal courts, but a designated individual must certify the record
Full impact card →US federal rules require one flight attendant per 50 seats above 100, seated near the exits for take-off and landing
Full impact card →Only the regulator and the certificated persons it names may approve an aircraft for return to service after maintenance, US aviation rules say
Full impact card →The maintenance record must carry the signature, certificate number and kind of certificate of the person approving the work, and that signature is the approval for return to service
Full impact card →Core work people keep#
The boundary a tool has to respect. Several of the stalled deployments above are what happens when a product crosses one of these.
- Explaining the numbers to decision-makers · Accountant / Bookkeeper
- Handling a customer who is already angry · Customer service representative
- Being responsible for the load · Truck driver
- Working out what the client actually needs · Graphic designer
- Judging quality and defending the call · Graphic designer
- Giving advice someone will act on · Lawyer
- Getting people to tell you things · Journalist
- Finding the message · Copywriter
- Understanding who actually buys and why · Marketing specialist
- Working out what is really being asked · Data analyst
- Knowing when the data is lying · Data analyst
- Deciding what to assume · Financial analyst
Show all 184
- Deciding whom to hire · HR / recruiter
- Handling difficult situations · HR / recruiter
- Running the room · School teacher
- Hands-on care · Registered nurse
- Talking to patients and families · Registered nurse
- Counselling the patient · Pharmacist
- Understanding the brief and finding the idea · Architect
- Defining the right problem · Product / UX designer
- Finding the story and the rhythm · Video editor
- Exploratory and adversarial testing · Software tester / QA engineer
- Gatekeeping and knowing what matters · Administrative assistant
- Helping with the complicated cases · Bank teller
- Restocking and shelf work · Retail cashier / shop assistant
- Dexterous or variable assembly · Assembly line worker
- Installation in real buildings · Electrician
- Cooking to order in a real kitchen · Chef / cook
- Running the team · Chef / cook
- The last fifty metres · Delivery rider / courier
- Handling the passenger · Ride-hail / taxi driver
- The conversation where it is decided · Sales / account manager
- The conversation that keeps someone · First-line manager / team supervisor
- After the plan breaks · Operations coordinator
- Deciding what to try first · AI implementation lead
- Deciding what it may do unsupervised · AI implementation lead
- Getting people to actually use it · AI implementation lead
- Agreeing what each side actually owes · Partnerships / channel manager
- Keeping them selling after the announcement · Partnerships / channel manager
- Changing a system you did not write · Experienced software engineer
- Designing for how it fails · Experienced software engineer
- The states nobody drew · Frontend developer
- Keeping it correct when things happen at once · Backend developer
- The data model, and changing it while it is in use · Backend developer
- Who is allowed to see what · Backend developer
- When the data is wrong and nothing errored · Data engineer
- Owning what a number means · Data engineer
- Deciding what not to build · Product manager
- Finding out what is actually wrong · Product manager
- Getting three teams to move together · Product manager
- Deciding what counts as good enough · Machine learning engineer
- Catching what nobody asked about · Radiologist
- Getting two sides to agree · Real estate agent
- Deciding where to look · Auditor
- Asking the question that gets a real answer · Auditor
- Signing it · Auditor
- The claim that does not fit · Insurance claims handler
- Telling someone no · Insurance claims handler
- Noticing what is not being said · Counsellor / psychotherapist
- The relationship itself · Counsellor / psychotherapist
- Making a therapeutic decision · Counsellor / psychotherapist
- The case the form was not written for · Government service clerk
- The person who cannot use the app · Government service clerk
- Negotiating with someone who will remember · Procurement / supply chain specialist
- Deciding who goes short · Procurement / supply chain specialist
- Reading what a result actually means · AI researcher
- Choosing what to work on at all · AI researcher
- Putting hands on the patient · General practitioner / primary care doctor
- Deciding it together, and saying the hard thing · General practitioner / primary care doctor
- The body work · Care worker / nursing assistant
- Being the person who is there · Care worker / nursing assistant
- Knowing a result is wrong · Medical laboratory technician
- Working out what is actually limiting them · Physiotherapist / rehabilitation therapist
- Treating with your hands · Physiotherapist / rehabilitation therapist
- Keeping them doing it · Physiotherapist / rehabilitation therapist
- Working out what they actually want · Retail salesperson / shop assistant
- Running the room · Waiter / restaurant server
- When it goes wrong · Waiter / restaurant server
- The awkward square metres · Cleaner / janitor
- Walking toward the situation · Security guard
- Finding out what actually happened · IT support specialist / helpdesk
- Deciding it is an incident · Security analyst (SOC)
- Being woken up · DevOps engineer / platform engineer / SRE
- Deciding how it should be built · DevOps engineer / platform engineer / SRE
- Actually using the thing you are documenting · Technical writer / documentation engineer
- Whatever walks in · Receptionist / front desk
- Finding the question behind the question · Management consultant
- Arguing the exception · Loan officer / credit officer
- Saying no to the business · Compliance officer
- Building it where it goes · Construction worker
- Working out what can happen today · Construction worker
- The fault that will not repeat · Auto mechanic / vehicle technician
- Doing the repair · Auto mechanic / vehicle technician
- Reading what the numbers mean · E-commerce operations specialist
- The platform changed the rules again · E-commerce operations specialist
- When the train stops in a tunnel · Metro train driver
- Lashing the boxes down · Container port worker
- The call the rule does not decide · Content moderator
- Turning a value into an enforceable line · Content moderator
- Keeping aircraft apart · Air traffic controller
- When the plan breaks · Air traffic controller
- Flying it by hand · Airline pilot
- When two systems disagree · Airline pilot
- The decision that is signed · Airline pilot
- The passenger in trouble · Bus driver
- Signing off on the numbers · Actuary
- Signing the plans as the qualified person · Civil / structural engineer
- Searching the building and getting people out · Firefighter
- Reading the fire and commanding the scene · Firefighter
- Being ready to evacuate the aircraft · Cabin crew
- Handling medical and security incidents · Cabin crew
- Responding to incidents and making arrests · Police officer
- Deciding what to plant, when and how · Farmer
- Working with clients over time · Social worker
- Assessing risk and responding to crises · Social worker
- Diagnosing and planning treatment · Dentist
- Carrying out treatment · Dentist
- Radiation safety and dose · Radiographer / radiologic technologist
- Installing and repairing pipes and fittings · Plumber
- Staying with the client over time · Financial adviser / financial planner
- Working out the cover a client needs and recommending it · Insurance agent
- Reviewing clients' cover as their lives change · Insurance agent
- Examining the animal · Veterinarian
- Treating, vaccinating and operating · Veterinarian
- Talking it through with the owner · Veterinarian
- Performing characters for animation and games · Voice actor
- Programmes, children and the community · Librarian
- Out-of-position, site and repair welding · Welder
- Difficult patients and unexpected findings · Sonographer
- Aligning stakeholders and negotiating trade-offs · Project manager
- Leading the team and clearing blockers · Project manager
- Deciding whether to operate, and the consent · Surgeon
- Operating by hand and through the laparoscope · Surgeon
- What the scans did not show · Surgeon
- Being the face and the voice · Content creator
- Framing the question and the metric · Data scientist
- Explaining what it means, and what it does not · Data scientist
- Understanding the client and the brief · Interior designer
- Requirements and system design · Electrical engineer
- Requirements and concept · Mechanical engineer
- Troubleshooting and repair · Aircraft maintenance technician
- Process design and improvement · Industrial engineer
- Portraits and events · Photographer
- Research and writing · University lecturer
- Character animation · Animator and VFX artist
- Court and legal interpreting · Interpreter
- Medical interpreting · Interpreter
- Performing live · Musician and composer
- Screenwriting and rewrites · Writer and author
- Principal screen performance · Actor and model
- Giving standardised tests · Psychologist
- Psychotherapy · Psychologist
- Medical nutrition therapy · Dietitian / nutritionist
- Signing out the diagnosis · Pathologist
- Selling and advising on complex trips · Travel agent / travel advisor
- Cleaning and scaling · Dental hygienist
- Installing systems · HVAC technician
- Hearing cases and running proceedings · Judge / magistrate
- Deciding the case · Judge / magistrate
- Lessons in real traffic · Driving instructor
- Airway management and emergencies · Anaesthesiologist / anaesthetist
- Cutting and styling · Hairdresser / barber
- Colouring and chemical treatments · Hairdresser / barber
- Trimming and pruning · Gardener / landscaper
- Checking age and serving alcohol responsibly · Bartender / barista
- Treating patients on scene · Paramedic / EMT
- Compliance, exceptions and corrections · Payroll specialist / payroll clerk
- Therapeutic and clinical massage · Massage therapist
- Preparing plans and policy · Urban planner
- Researching and determining boundaries · Land surveyor
- Designing surveys, samples and trials · Statistician
- Explaining and witnessing signatures · Notary
- Drafting deeds (civil-law notaries) · Notary
- Airway care and breathing treatments · Respiratory therapist
- Dispatching and coordinating units · Emergency dispatcher
- Giving instructions until help arrives · Emergency dispatcher
- Setting up machines · Machinist / CNC machinist
- Assessing communication and swallowing · Speech-language pathologist / speech and language therapist
- Monitoring labour · Midwife
- Attending births and emergencies · Midwife
- Planning and delivering therapy · Occupational therapist
- Caring for patients during the draw · Phlebotomist
- Policy analysis and advice · Economist
- Assisting at the chair · Dental assistant / dental nurse