If artificial intelligence really is coming for entry-level jobs, perhaps we should first send it after the plugged toilets.
That is the thinking behind T-Rex, my hypothetical idea for an AI-controlled commercial restroom-cleaning robot. T-Rex is not a Veristio product. It is simply a useful thought experiment. Restroom cleaning can be unpleasant, repetitive, hazardous, and chronically understaffed. If a machine could handle much of that work, would we say it had destroyed an entry-level job? Or would we ask who operates it, checks its work, replaces a failed part, handles an exception, documents a problem, and decides when the restroom is actually safe?
That second question gets closer to what is happening across the economy. AI and automation are eliminating some positions and shrinking some kinds of work. We should not minimize that. But they are also changing the tasks inside jobs and moving the first rung of the employment ladder.
What entry level used to mean
Entry-level work traditionally meant work a person could begin with limited occupational experience. The job itself helped teach the industry. A new employee sorted documents, entered information, answered routine questions, moved material, inspected simple defects, or watched an experienced worker solve problems.
Those beginner tasks did more than produce output. They gave people context. A clerk learned how a business handled unusual cases. A junior analyst learned which numbers could be trusted. A warehouse worker learned how goods actually moved. An assistant learned the difference between a written procedure and what happened when the procedure met reality.
This is why the distinction among a task, a job, and an occupation matters. A task is one piece of work. A job is the bundle of tasks performed by one person for an employer. An occupation groups similar jobs across employers. Technology may automate several tasks without eliminating the whole job. It may also remove enough beginner tasks that an occupation survives while its traditional entrance narrows.
The International Labour Organization’s 2025 assessment examined nearly 30,000 tasks and concluded that one worker in four worldwide is in an occupation with some exposure to generative AI. Its more important conclusion was that transformation is more likely than complete redundancy because most occupations still require human input. Exposure is not a layoff forecast. It is a signal that the contents of work may change.
The easy pile is disappearing
Automation is especially good at volume, repetition, pattern matching, routing, and consistent physical motion. Software can classify routine messages, extract standard fields from documents, draft ordinary summaries, and answer familiar questions. Machines can space, route, scan, label, carry, sort, and reject items at speeds people cannot match.
I have seen versions of this change firsthand. At a Lowe’s regional distribution facility in Cheyenne, Wyoming, I observed extensive automation, robots, conveyor-spacing systems, routing equipment, and pass-through label readers. At a food-canning plant in Moline, Illinois, I saw industrial vehicles following designated floor paths. They may have used magnetic guidance, but I did not verify the mechanism. At Coors operations in Golden, Colorado, high-speed conveyors moved empty cans while air jets rejected defective ones.
These observations do not prove what every facility does. They do show the practical shape of automation. The machine handles a repeatable flow. People build, monitor, adjust, inspect, clean, maintain, and recover that flow when reality refuses to stay repeatable.
An OECD review of workplace case studies found this same pattern in office work. Tools handled routine document extraction, basic customer inquiries, and easy insurance cases. Employees were left with the exceptions and more complex judgments. That can raise productivity, but it creates a career-ladder problem. The easy pile was often where beginners learned.
If the easy pile disappears, employers may ask new hires to handle exceptions before they have seen enough ordinary cases to recognize one. A company can keep the occupation while weakening the apprenticeship hidden inside it. The danger is not only fewer jobs. It is fewer safe ways to become competent.
The first rung is moving
The new first rung is likely to appear around automated systems rather than underneath them. It includes operating equipment, checking outputs, responding to alarms, documenting failures, performing preventive maintenance, escalating unusual cases, testing corrections, and explaining results to customers or supervisors.
These roles are not automatically glamorous, highly paid, or plentiful. Some will be better jobs. Some will be more stressful because workers receive only the difficult cases. Some employers will use technology to intensify work or monitor employees too closely. The gains and losses will not be shared evenly.
Still, the surrounding work is real. The Bureau of Labor Statistics describes industrial machinery mechanics as workers who install, maintain, and repair production equipment. O*NET lists tasks such as repairing machinery, diagnosing malfunctions, and replacing failed components. Those responsibilities exist because automated equipment does not maintain itself.
The same logic applies outside factories. A person using an AI system must know when the source material is incomplete, when an answer conflicts with policy, when a customer needs human attention, and when the cost of being wrong is too high. Judgment, accountability, communication, verification, maintenance, and exception handling are not decorative additions. They are the work that keeps automation connected to reality.
Preparation without a computer-science degree
Useful preparation does not require turning every student into a software engineer. Most people do not need to build an AI model. They need to work competently with systems that contain software, sensors, data, machinery, and rules.
That requires several practical abilities:
First, learn the process, not merely the interface. A person who understands how a warehouse, clinic, office, plant, or repair operation creates value can recognize when a tool is producing nonsense.
Second, practice verification. New workers should learn to check sources, measurements, tolerances, records, and outputs. “The computer said so” is not a quality standard.
Third, learn basic troubleshooting. That may mean reading an error code, isolating a bad sensor, comparing expected and actual results, documenting a repeatable failure, or knowing when to stop the system and call someone with deeper expertise.
Fourth, strengthen communication. Exception-heavy work requires clear notes, useful questions, calm explanations, and honest escalation. A worker who can explain what happened and what has already been tested saves time and prevents repeated mistakes.
Fifth, understand responsibility. Automated recommendations still operate inside organizations governed by safety rules, contracts, professional duties, and law. Someone must remain answerable for consequential decisions.
Schools and training providers should build these abilities into real projects. Community colleges can combine industrial maintenance, electronics, logistics, health technology, business operations, and data literacy. Technical programs can teach sensors and controls alongside safety and documentation. Certificates can demonstrate a bounded skill without pretending to replace experience. Apprenticeships can preserve guided learning when routine tasks no longer provide it automatically.
Employers also have obligations. They cannot automate the training tasks, demand experienced judgment from every applicant, and then complain about a skills shortage. They should create supervised entry roles, simulations, shadowing, equipment labs, paid apprenticeships, and clear progression from operator to technician or specialist. Experienced employees need time and recognition for teaching.
Tool-assisted learning can help. AI can generate practice scenarios, explain unfamiliar terms, and provide low-cost repetition. But learners should also be shown where the tool fails. Training that teaches only prompting produces dependence. Training that combines tools with domain knowledge, verification, and hands-on work produces capability.
A ladder must still touch the ground
The optimistic mistake is to assume every displaced worker will smoothly step into a better technical role. The pessimistic mistake is to assume the ladder simply disappears.
Some traditional entry-level positions will vanish. Others will contract. Many will be rebuilt around harder tasks. New roles will emerge, but they may appear in different places, require different preparation, or remain inaccessible to people who cannot afford unpaid retraining. That is why the transition cannot be left to slogans.
Students and workers should look beyond job titles and ask what systems are entering an industry, what those systems cannot reliably do, and who is responsible when they fail. Schools should teach process knowledge and verification. Employers should rebuild the learning pathways their automation removes. Policymakers and workforce institutions should support apprenticeships, community colleges, short technical credentials, and paid transitions tied to actual regional employers.
AI is not simply erasing entry-level work. It is relocating the beginning. Our challenge is to make sure the new first rung is visible, teachable, paid, and low enough for a beginner to reach. A modern ladder can be made of sensors, software, machines, and human judgment. But it is useless if it no longer touches the ground.