Beyond estimating, electricians use AI to explain inspection and test findings to customers in plain English, cross-check product datasheets, tidy circuit charts, write variation emails to builders, pre-qualify EV charger and solar enquiries, log intermittent faults, produce handover packs, fill in pre-qualification forms and revise regulation changes. Every technical decision stays with the electrician.
The common thread: electrical work generates more explaining and documenting than most trades. Customers don't understand the report, main contractors want everything in writing, and product manuals run to 80 pages. AI is good at the words around the work. It is not good at, and should never be trusted with, the calculations, measurements and judgements that make the work safe.
1. Explaining inspection and test findings in plain English
What it does: turns the observations on a periodic inspection report into a short explanation a homeowner or landlord can act on. The formal report stays exactly as you wrote it; the plain-English version goes in the covering email.
Example, from a small testing firm that works for landlords (illustrative). The electrician pastes three observations with their codes, in the wording from the report, and asks: "Explain these to a landlord who isn't technical, in order of urgency, one short paragraph each, no scare language, and say what we recommend." The draft explains that the first item means part of the installation lacks the extra protection modern circuits have against electric shock, and why it matters in a rented property; that the second is a damaged socket front in the kitchen that should be replaced soon; and that the third is a labelling issue that is advisory. The electrician then checks every explanation matches his own judgement, corrects one sentence that downplayed the urgency of the first item, and adds the price for each remedial job.
How to start: save a prompt that includes your firm's tone and the phrase "do not change the urgency or meaning of any observation".
Watch out: AI softens. It will turn "requires urgent attention" into "worth looking at when convenient" unless told not to. The code and its meaning are yours; the model only rephrases.
2. Cross-checking datasheets before you order
What it does: reads two or three product datasheets side by side and answers specific compatibility questions, with page references you can check.
Example, from a domestic electrician (illustrative). A customer has chosen LED downlights and a particular smart dimmer. The electrician uploads the dimmer's datasheet and the downlight's specification to Gemini Notebook (formerly NotebookLM), which answers from the documents you give it and shows where each answer comes from. Asked "Is this dimmer listed as compatible with this lamp type, what's the minimum load, and how many of these lamps can it run?", it quotes the dimmer's minimum load and dimming type, and points out that the downlight sheet lists a different dimming method from the one the dimmer supports. That mismatch is exactly the kind of thing that produces flicker complaints a month after the job.
How to start: keep a notebook per manufacturer you install often, with current datasheets and installation manuals. A setup that works in an afternoon:
- Download the current datasheets and installation manuals from each manufacturer's website, not from a wholesaler's listing, which can lag behind.
- Rename each file with the product code and the date you downloaded it, so an answer's source is obvious at a glance.
- Create one notebook per manufacturer, or per product family you fit often (dimmers and lamps, EV chargers, smart controls).
- Test it with three questions you already know the answers to. If any answer is wrong or uncited, check the file uploaded properly before trusting it with new questions.
- Every six months, replace old datasheets with current ones and delete the old versions.
Watch out: the answer is only as good as the documents. Ask a general chatbot the same question without datasheets and it may answer confidently from a different product. Always open the referenced page and read it yourself before ordering. For anything involving load totals, do the arithmetic yourself.
3. Tidying circuit charts and labels
What it does: turns your notes, or a photo of a scribbled chart, into a clean, consistently worded circuit schedule you can print and fix inside the board cover.
Example (illustrative). After a board change in a three-bedroom house, the electrician dictates: "way one, cooker. two, sockets kitchen and utility. three, sockets downstairs front. four, upstairs sockets. five, lights down. six, lights up plus loft. seven, shower. eight, garage feed." The assistant returns a two-column table with ways numbered, descriptions in a consistent format ("Sockets: kitchen and utility"), and a note line at the bottom for the date and the firm's name. The electrician adds protective device ratings from his own test sheet, not from the AI.
How to start: give it your preferred label format once and reuse the prompt.
Watch out: photos of handwritten charts are misread often, especially numbers. Check every way against the board before fixing the chart. Never let the model "fill in" device ratings or cable sizes it can't know.
4. Writing variations and change emails to builders
What it does: turns "the site manager asked us to move six sockets" into a clear written variation with what changed, why, the cost, and the effect on the programme, sent the same day.
Example, from a three-person commercial fit-out contractor (illustrative). Notes: "Site manager asked verbally Tuesday to relocate 6 double sockets in office 2 after partition moved. Extra 1 day for 2 men plus materials. Need written instruction before we start." The draft email to the main contractor sets out the instruction received (date, who gave it), the scope of the change, the cost from the firm's day rates and materials list, the extra day, and a closing line: "Please confirm by written instruction so we can proceed; we'll hold this work until then." The electrician checks the cost against his own sums and sends it.
How to start: keep a variation template and paste your rates into the prompt so figures come from your list.
Watch out: contract terms. If your subcontract has a specific notice procedure or deadline for variations, put that in the prompt, and don't let the draft agree to anything the contract doesn't say. Builders deal with the same problem from the other side; how small builders price variations with AI is useful reading for both.
5. Pre-qualifying EV charger, solar and battery enquiries
What it does: drafts the questions a customer should answer, and the photos they should send, before you visit, then sorts the replies into "survey needed", "probably straightforward" and "unlikely to be suitable, call first".
Example, from an EV charger installer (illustrative). The installer asks AI to turn his survey checklist into a friendly web form: where the car parks, distance from the fuse board to the parking spot, whether the property has off-street parking, photos of the fuse board with the cover on, the meter, and the route a cable might take. When replies arrive, he pastes each one in and asks for a two-line summary and a list of anything missing. One reply shows a board that looks old and crowded; the summary flags "board photo suggests a survey is needed before quoting", which is his call to confirm.
How to start: build the question list from your last ten surveys: what did you wish you'd known before driving out?
Watch out: suitability, supply capacity and earthing arrangements are judged on site by you, not from a photo by a model. Use the sort to plan visits, never to quote a fixed price.
6. Logging intermittent faults and handing them over
What it does: keeps a clean log of what has been tested, when, and with what result, so the second visit (or a colleague) doesn't repeat the first.
Example (illustrative). An RCD in a customer's house trips at random, two or three times a week. After each visit, the electrician dictates what he did. Before AI, the notes were fragments across three text messages. After, the assistant maintains a single log:
Fault log - [job ref] - RCD 1 random trips
Visit 1 (Mon): insulation resistance tests on circuits 1-4, all
satisfactory. Customer reports trips often after rain.
Visit 2 (Thu): outside socket and garden lights disconnected for trial.
Customer to report any trips.
Open questions: trips linked to rain? Any trips since Thu?
Next step (electrician's decision): [ ]The pattern the log makes visible, trips after rain, came from the customer's comment on visit 1 that would otherwise have been lost. The electrician decides what that means and what to test next; the log makes sure nobody forgets it.
How to start: one running note per fault job, updated by voice after each visit.
Watch out: if you ask AI for "likely causes", treat the answer as a reminder list, not a diagnosis. It can't see the installation.
7. Customer handover packs
What it does: produces a one- or two-page handover for the customer: what was installed, where things are, what the test buttons are for, what to do if something trips, and who to call.
Example, from a domestic electrician after a board change and smart heating install (illustrative). The prompt includes the job sheet, the product names and the customer's questions from the day ("what do I do if the power goes off?"). The draft covers where the board is, what each section does in plain words, how to reset a tripped device safely, why and how often to press test buttons (from the manufacturer's instructions the electrician pasted in), how to use the heating app, and warranty details. The customer gets it by email the same evening, and the electrician gets fewer "quick question" calls the following week.
How to start: make one good handover by hand, then use it as the model in every prompt.
Watch out: reset instructions must be right and safe. Check that section every time; don't let the model add advice beyond what you'd tell the customer yourself.
A realistic slip shows why (illustrative). A handover pack for a new board told the customer that if the power went off, they should "reset the main switch". The job sheet pasted into the prompt listed the devices, but the model assumed a common layout rather than this board's. The customer's actual problem, a tripped RCD protecting the kitchen circuits, wasn't the main switch at all, and she rang the electrician in a panic on a Sunday. The corrected pack describes this board: which device protects what, what a tripped one looks like, and the reset steps the electrician would give in person. Board-specific wording, checked against the board, is the fix.
8. Pre-qualification forms and paperwork for commercial work
What it does: helps answer the pre-qualification questionnaires, supplier forms and document requests that come with commercial work, drawing on your own policies, insurance details and past projects.
Example, from a small commercial contractor (illustrative). A facilities company sends a 40-question supplier form. The contractor keeps a folder with his health and safety policy, insurance summary, training records and five project descriptions. Uploading those to a project in his chat assistant, he asks it to draft answers to each question using only those documents and to mark any question it can't answer from them. It drafts 31 answers and flags nine, including one about carbon reporting the firm has never been asked before. He writes those himself and checks every drafted answer against the source. What used to take a full evening takes about two hours.
How to start: gather your standard documents into one place first; that alone speeds up every form. For larger bids, using AI to respond to tenders covers the process.
Watch out: never let a drafted answer claim a certification, accreditation or policy you don't hold. The "use only these documents" instruction and your read-through are what prevent it. Risk assessments and method statements that a client requests are drafts until someone competent has checked them against the actual site.
9. Keeping up with regulation changes and training apprentices
What it does: summarises official guidance and amendments you've obtained, and turns your own notes into practice questions for apprentices.
Example (illustrative). An electrician with two apprentices loads the official guidance documents his firm has bought, plus his own training notes, into a notebook. He asks for ten practice questions on safe isolation and inspection sequence, each with the answer and where it comes from in the documents. He reviews the questions before the apprentices see them, removes one that was ambiguous, and uses the rest in a Friday afternoon session. For himself, he asks the notebook to list what changed in the latest amendment document, with section references, and reads those sections properly.
How to start: only use the official documents you're entitled to, and keep the notebook to those sources.
Watch out: this is where general chatbots are most dangerous. Asked from memory, they misquote regulation numbers, mix up editions and blend rules from different countries. Never take a regulation, value or limit from AI without the source open in front of you.
Why estimating isn't on this list
Estimating and take-offs are the AI uses electricians hear about most, and they deserve their own treatment. Counting fittings from drawings is covered in how electricians use AI to count fittings from drawings, and dedicated tools are compared in the best AI estimating software for small electrical firms. The nine uses above are the ones that eat the hours around every job, estimated or not. For the general trade jobs that apply to electricians too, such as enquiry replies, invoice chasers and review requests, see what tradespeople actually use AI for.
The nine at a glance: what to check each time
Each use has a different failure mode, so each needs a different check. Pin this next to your desk until the checks become habit:
| Use | Typical time on a job (illustrative) | The check that matters |
|---|---|---|
| 1. Test findings in plain English | 20 minutes down to 5 | Urgency of every observation unchanged |
| 2. Datasheet cross-checks | 30 minutes of scrolling down to 5 | Open the cited page and read it |
| 3. Circuit charts | 15 minutes down to 5 | Every way checked against the board |
| 4. Variation emails | 25 minutes down to 8 | Cost and contract procedure |
| 5. Enquiry pre-qualification | One wasted visit avoided now and then | No suitability verdict from photos |
| 6. Fault logs | Minutes, but a repeat visit saved | Log matches what you actually tested |
| 7. Handover packs | 30 minutes down to 10 | Reset and safety instructions correct |
| 8. Pre-qualification forms | An evening down to about 2 hours | No claimed accreditation you don't hold |
| 9. Regulation revision | Faster reading, not less reading | Source open; never from memory |
Two patterns stand out. The biggest time savings are on writing jobs where the facts are already yours (findings, handovers, forms). The biggest value, though, often comes from the small ones that prevent a return visit: a dimmer mismatch caught before ordering, or a fault log that stops a colleague repeating yesterday's tests. Those rarely show up in a time count, so note them separately when you review how the month went.
A week at a two-person domestic firm using four of these
To see how the nine combine, picture a two-person domestic electrical firm: the owner does surveys, testing and paperwork; an improver does most installs. A plausible week once four of these uses are in place (illustrative):
- Monday: two periodic inspections for a landlord. The owner writes the formal reports as usual, then applies use 1 to produce a one-page plain-English summary for each, with remedial prices. Before, the landlord often rang to ask what the codes meant; this week, he approves one remedial job by email the same afternoon.
- Tuesday: a customer wants smart dimmers with the downlights she has already bought. Use 2 flags that her lamps list a different dimming method from the dimmer she chose. The owner suggests a compatible dimmer before ordering, avoiding a return visit.
- Wednesday: a board change. The improver dictates the circuits; use 3 produces the chart, and the owner adds device ratings from his test sheet.
- Thursday: use 7 turns Wednesday's job sheet into a handover pack, checked against the board and sent that evening.
- Friday: the week's variation from a small builder they subcontract for, written with use 4 and sent before the weekend, so the extra work is agreed before anyone starts it on Monday.
None of that replaced any electrical work. It moved perhaps three or four hours of evening writing into short, checked tasks during the day, and headed off one return visit and one unpaid variation, which are often worth more than the hours.
Where electricians should keep AI out entirely
- Design calculations: cable sizing, voltage drop, maximum demand, fault current, disconnection times. AI makes arithmetic errors and states them confidently.
- Test values and limits. Take them from your instruments and the official tables, never a model's memory.
- Certification and sign-off. Every certificate you sign reflects your inspection and your competence.
- Safe isolation decisions on site. A checklist helps you remember; it doesn't replace proving dead.
- Anything you can't check. If you couldn't verify the answer against a document or your own knowledge, don't use it.
What these nine cost, and a sensible order to try them
All nine run on a general chat assistant, typically about $20 a month for a paid plan such as ChatGPT Plus, Claude Pro or Google AI Pro; datasheet and regulation work suits a notebook-style tool that answers only from uploaded documents. Some job-management apps include AI writing features that cover handover and variation text inside the job itself; check yours before adding a subscription.
A sensible order for a small firm: start with test-finding explanations (1) and handover packs (7), because customers notice them immediately and they reuse the same prompt every time. Add datasheet checks (2) next, since one avoided flicker callback pays for months of the tool. Leave pre-qualification forms (8) until your standard documents are in one folder; the folder is half the saving.
A quick check that it's working: after a month, count the "quick question" calls and emails from customers after jobs, and the time spent on each variation or form. If neither has moved, the prompts need more of your own examples in them, not a different tool.
Further reads
- Why Is AI Bad at Maths? What to Check in Quotes and Invoices — Why AI must never do your cable or load calculations.
- Seven AI Mistakes That Cost Tradespeople Jobs — The AI errors that lose trades businesses work.
- The First 30 Days of AI in a Trades Business — A practical first month of AI in a trades firm.
- Compliance Calendar: Track Licences, Insurance and Certificates — Keep registrations and insurance renewals on track.
- Replying to Trade Enquiries in 60 Seconds With AI — Fast enquiry replies from the van.
- Getting More Reviews After Every Job With AI Follow-Ups — Getting reviews after every job, automatically.
- When Should a Trades Business Bring In an AI Consultant? — When it makes sense to bring in outside help.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Google help page on Gemini Notebook (formerly NotebookLM); manufacturer datasheet and installation manual conventions; general guidance on AI arithmetic and document grounding.