Will AI replace plumbers, electricians and dentists: what is changing in jobs done with your hands
Topics: AI, Small business, Automation

The short answer: based on the data available now, AI does not do the main physical work in these professions, and the first robots in dentistry work to a plan that a dentist approves. In studies by Microsoft and Anthropic of how people use chatbots, manual work ended up at the very bottom of the list. The US Bureau of Labor Statistics expects more electricians, plumbers and dentists by 2035, and for electricians one of the reasons it names is rising electricity demand driven by AI itself and data centers. What changes are the tasks that surround the physical work: tradespeople hand requests, client scheduling, estimates and paperwork over to software, and a dentist gets a second opinion on an X-ray from a program. Below we look at what already works in each profession, where AI reaches its limits and where the owner of a service company or clinic should start.
What research says about manual work and AI
In a 2025 paper, Microsoft researchers analyzed 200,000 anonymized conversations with the Bing Copilot chatbot and matched them to the work tasks of different occupations. The finding: AI most often helps to find, process and summarize information. At the bottom of their list of occupation groups were construction workers and their helpers, plant and system operators (for example, at water treatment plants), home health aides and orderlies, and agricultural and forest workers.
In February 2025, Anthropic ran a similar analysis of about a million conversations with Claude. Occupations with a large share of physical labor appeared there least often: for example, farming, fishing and forestry accounted for 0.1% of queries.
It matters what exactly these studies measured. They show which tasks people bring to a chatbot, not who will be laid off. The Microsoft authors explicitly warn that drawing conclusions about job losses from their table would be a mistake, and give the example of ATMs: they took over part of bank tellers' work, yet the number of tellers in US banks grew afterwards.
Demand for tradespeople and dentists is growing
The US Bureau of Labor Statistics (BLS) publishes ten-year projections by occupation. The latest projection, for 2025-2035:
| Occupation | Employment growth | Openings per year on average | Reason for demand according to BLS |
|---|---|---|---|
| Electricians | 9% | about 72,700 | wiring in nearly every building, connecting solar and wind power, rising electricity demand from AI and data centers |
| Plumbers and pipefitters | 7% | about 42,000 | new construction, repair of systems in existing buildings, mandatory fire suppression systems |
| Dentists | 6% | about 4,800 | more older people, people keeping their own teeth longer, demand for cosmetic dentistry |
BLS describes growth of 9% and 6% as above the average for all occupations. This is a projection for the US, not a guarantee, and it cannot be applied to Russia directly.
In Russia, demand for tradespeople shows in job postings. According to hh.ru data cited by the outlet MSK1, in March 2026 there were more than 400 openings for electrical installers in St. Petersburg alone, and about as many for plumbers. The median salary in St. Petersburg in the first quarter of 2026 was 99,100 rubles for electricians (up from 92,600) and 90,300 rubles for plumbers (up from 88,300). Tradespeople the outlet spoke to put the wait for routine jobs at two to three weeks.
Plumbers and electricians: AI takes the paperwork, the wrench stays with the tradesperson
Where AI already helps tradespeople can be seen in a survey by Housecall Pro. It is an American company that sells software for service businesses, so its figures should be read with the seller's interest in mind. In June 2025 it surveyed more than 400 owners and tradespeople in the US, including plumbers, electricians and heating and air conditioning technicians:
- more than 70% had at least tried AI tools, about 40% use them in their work;
- those who use AI, by their own estimate in the survey, freed up more than four hours a week from administrative work: emails, data entry, reminders to clients. This is a self-assessment, not a measurement;
- AI is used for advertising, replying to clients, scheduling and paperwork;
- 57% of users believe AI helped their business grow, mainly through faster replies to requests.
Why the call-out itself stays with a person is explained by the same Microsoft data: the strength of today's AI is working with information, while a tradesperson's work on a call-out is physical: pipes, wiring and the electrical panel in someone else's home. None of the sources for this article describe a robot that does this kind of work on call-outs. This is a conclusion from the data above, not a measurement.
If you have a leak at home or a sparking socket and want to ask a chatbot for advice, the line is simple.
- Water. A chatbot can help you figure out where to shut off the valve on the riser or at the water inlet to your flat, and how to describe the fault to a plumber. Only attempt a repair yourself based on its advice once the water is shut off and you understand what you are doing. Otherwise, it is better to wait for a plumber.
- Electricity. If a socket or wiring is heating up, sparking or smells of burning, while the electrical panel itself is fine (not sparking, not smoking, not hot) and you can reach it without touching the damaged socket, wiring or wet areas, you can switch off the power with the circuit breaker in the panel and call an electrician. If the panel itself is faulty or you are not sure it is safe to approach, do not touch anything in it: step away and call your building management company's emergency service or an electrician. If there is smoke or fire, call 112. Do not open the panel or change wiring following a chatbot's instructions: a mistake here risks electric shock and fire.
Dentists: AI looks at the X-ray, the dentist decides
In dentistry, AI already works inside the clinic. The most visible use in the sources for this article is finding signs of disease on X-rays.
Software for X-rays. In March 2022, the US regulator FDA cleared for the first time a program that reads dental X-rays: Second Opinion by Pearl. It marks suspected caries, tartar, inflammation at the root and other findings on the image. According to the company itself, in its studies dentists assisted by the program found 36% more lesions than without it. In December 2025, the FDA extended the clearance to panoramic X-rays.
How reliable such programs are was assessed by the authors of a review in the journal Caries Research (2025). They examined 23 studies of neural networks that look for caries on bitewing X-rays: such an X-ray shows the crowns of several upper and lower back teeth at once. The conclusion is cautious: the results are promising, especially for deep lesions, but every study has a high risk of bias in at least one quality criterion, and there is no testing on X-rays from other clinics. Put simply, the programs show good results in studies, but how they will perform in a particular clinic has to be checked.
Similar systems exist in Russia too. For example, in April 2026 ComNews reported on a pilot project of the system "Third Opinion" (Tretye Mneniye) with the La Peregrina clinic chain: the program analyzes three-dimensional jaw scans (CBCT, cone beam computed tomography) and marks caries, inflammation of the tissue at the tooth root, unerupted teeth and previous treatment. The article stresses that AI is used not instead of the dentist but to analyze complex scans.
What the law requires. In Russia, a program that automatically analyzes medical data and influences a doctor's decision is considered a medical device (Article 38 of Federal Law No. 323-FZ on the fundamentals of health protection; the criteria are set out in a letter from Roszdravnadzor, the federal healthcare oversight agency, No. 02И-297/20 of 13.02.2020). As a general rule, such a device may only be used with a registration certificate. In August 2026, Sergei Kolesnikov, Deputy Minister of Health of the Nizhny Novgorod Region, gave the number of registered AI solutions for medicine: 57.
Robots. Here too there are real examples, but with caveats.
- The Yomi robot by Neocis received FDA clearance in 2017 as a system for placing implants. It does not work on its own: the surgeon guides the drill, the robot holds it to a plan prepared in advance, and the surgeon is in control throughout.
- In summer 2024, the company Perceptive reported the first fully automated dental procedure on a human. The robot prepared a tooth for a crown in Colombia. The company claims such a procedure takes 15 minutes instead of two visits. At the time of the procedure the system had no FDA approval, and the robot's work plan is approved by a dentist before it starts.
So in dentistry AI already helps the dentist see and plan, but the diagnosis, the treatment plan and the responsibility remain with the dentist.
Where the owner of a service company or clinic should start
Below is an order that follows from the data above: first, whatever takes up people's time and does not require hands.
- Requests and scheduling. For a service company, a missed call can mean a lost client. You can start with an automatic call answering system or a messenger bot that takes requests around the clock, asks for the address and the nature of the fault and offers a time slot. Who does it: an automation contractor, or the owner using a ready-made service. How to check: count missed calls and requests for the month before launch and the month after.
- Estimates and documents. The tradesperson dictates or writes down what they saw on site, and the program puts together a draft estimate and a draft letter to the client. The tradesperson checks prices and the scope of work personally: a chatbot can be confidently wrong. Who does it: the tradesperson using a ready-made service; the owner sets up the estimate template once. How to check: the time from the site visit to sending the estimate to the client.
- Reminders and repeat visits. For a clinic, these are reminders about appointments and check-ups; for a service company, an offer of a scheduled inspection of the boiler or the electrical panel. Who does it: the administrator sets up the messages in the scheduling software. How to check: the share of missed appointments or of repeat orders before and after.
- AI for X-rays in a clinic. Who does it: the chief physician or the clinic owner. Before buying, ask the supplier for the registration certificate number and look it up in the state register of medical devices on the Roszdravnadzor website. No certificate: do not use the program with patients. For the first few weeks, compare the program's findings with the dentist's conclusion to see how it performs on your clinic's own X-rays.
- Client data. Do not paste addresses, phone numbers or patients' medical information into a public chatbot. To work with them you need a service you have a contract with and where it is clear where the data is stored. The Russian personal data law (152-FZ) requires databases with data on Russian citizens to be stored in Russia.
How to tell whether it paid off. The figures from the steps above show the benefit, but it has to be compared with the costs. Monthly costs: the service subscription, plus the one-off setup divided by the number of months you plan to use it, plus staff time spent looking after it. Monthly benefit: additional orders multiplied by the average profit per order, plus freed-up hours multiplied by the cost of an employee's hour, but only if those hours went to other paid work or allowed you to cut costs. Otherwise the calculation will overstate the benefit. A hypothetical example: if the service costs 5,000 rubles a month and the profit from one call-out is 2,000 rubles, it pays off when it brings more than two additional call-outs a month. If after two or three months the benefit does not cover the costs, the service should be reconfigured or switched off.
If the company has one or two tradespeople and few requests, an ordinary calendar and message templates may be enough. If there are almost no missed requests and estimates reach clients on the day of the visit, there is nothing to invest in automation for yet.
Summary
AI is currently changing manual professions differently from office jobs: it does not replace the hands of the tradesperson or the dentist and takes over part of the paperwork and information work around them. For plumbers and electricians this means requests, estimates and correspondence; for dentists, also a second opinion on X-rays, while the diagnosis and the responsibility remain with the dentist. Demand projections for these professions in the US are rising, and in St. Petersburg, according to tradespeople, the wait for routine jobs reaches two to three weeks. It makes sense for an owner to start with whatever eats up people's time and to measure the result by missed requests, time to estimate and missed appointments.
I work on AI agents and automation. If you would like to see my projects or discuss your own task, take a look at my portfolio.
Sources
- Tomlinson et al., Microsoft Research, 2025: "Working with AI: Measuring the Applicability of Generative AI to Occupations". https://arxiv.org/abs/2507.07935
- Anthropic, 10.02.2025: The Anthropic Economic Index. https://www.anthropic.com/news/the-anthropic-economic-index
- U.S. Bureau of Labor Statistics, Occupational Outlook Handbook: Electricians. https://www.bls.gov/ooh/construction-and-extraction/electricians.htm
- U.S. Bureau of Labor Statistics, Occupational Outlook Handbook: Plumbers, Pipefitters, and Steamfitters. https://www.bls.gov/ooh/construction-and-extraction/plumbers-pipefitters-and-steamfitters.htm
- U.S. Bureau of Labor Statistics, Occupational Outlook Handbook: Dentists. https://www.bls.gov/ooh/healthcare/dentists.htm
- MSK1, 12.04.2026: why it is so hard to find a plumber or an electrician in Russia (in Russian). https://msk1.ru/text/job/2026/04/12/76360471/
- Plumber Magazine, June 2025: Housecall Pro survey on AI among home service professionals. https://www.plumbermag.com/online_exclusives/2025/06/survey-finds-many-home-service-pros-using-ai-tools-to-streamline-admin-work
- Pearl, 08.03.2022: FDA clears the world's first AI software to read dental X-rays. https://hellopearl.com/news/fda-clears-worlds-first-ai-software-to-read-dental-x-rays-3e780
- Pearl, 18.12.2025: FDA clearance for panoramic X-rays. https://hellopearl.com/news/pearl-expands-dental-ai-capabilities-with-fda-clearance-for-panoramic-x-rays
- Hansson Horvath, Gjerdet, Shi, Caries Research, 2025: review of deep learning for caries detection on bitewing X-rays. https://pmc.ncbi.nlm.nih.gov/articles/PMC12331221/
- ComNews, 17.04.2026: AI improves diagnostics and planning in dentistry (in Russian). https://www.comnews.ru/digital-economy/content/244852/2026-04-17/2026-w16/1012/ii-uluchshaet-diagnostiku-i-planirovanie-stomatologii
- Brace law firm, 24.03.2026: software as a medical device, registration criteria (Article 38 of 323-FZ, Roszdravnadzor letter No. 02И-297/20; in Russian). https://brace-lf.com/informaciya/farmatsevticheskoe-i-meditsinskoe-pravo/programmnoe-obespechenie-meditsinskoe-izdelie-kriterii-registratsii
- EAN, 28.08.2026: the regional health ministry on how doctors use AI (in Russian). https://eanews.ru/rossiya/20260828143107/v-minzdrave-rasskazali-kak-vrachi-ispolzuyut-ii
- MassDevice, 2017: Neocis wins FDA clearance for the first dental robotic guidance system. https://www.massdevice.com/neocis-wins-fda-nod-1st-dental-robotic-guidance-system/
- Popular Science, 2024: a robot dentist prepared a tooth for a crown. https://www.popsci.com/technology/robot-dentist/
- BGR, 06.08.2024: the robot performed a fully automated procedure in Colombia; FDA approval about five years away. https://bgr.com/science/a-robot-just-performed-fully-autonomous-surgery-on-a-live-patient-for-the-first-time/
- Dentistry.co.uk, 01.08.2024: "Robot dentist" completes first dental procedure. https://dentistry.co.uk/2024/08/01/robot-dentist-completes-first-dental-procedure/