ChatGPT for tradespeople: tasks you can hand over today
Written with AI and checked against the sources by a second AI. As of November 15, 2026. More in the notes on AI.
Writing quotes, typing up notes from a site survey, asking customers for an appointment: this often eats into your evening. You can hand much of it to ChatGPT or another AI, as long as you tell it exactly what to deliver and check the result before it goes out.

Key points
- ChatGPT phrases, sorts and calculates by your rules. Facts, measurements and prices come from you.
- It works well for quote texts, customer e-mails, material lists from notes, daily reports, checklists and small calculators.
- Don’t hand over structural calculations, standards, safety-related values or prices you have not checked yourself.
- Your customers’ names, addresses and phone numbers don’t belong in a personal AI account. Use placeholders.
- Check every result against your notes, your costing or a job you already know.
Which tasks fit
An AI like ChatGPT is good at turning your bullet points into a tidy text, sorting a muddle into a table and calculating by fixed rules. But it has never been to your site. It knows neither the house nor the customer nor your hourly rate. Everything it needs to know about them is in your instruction, the text you send it.
| Task | What you give it | What you check |
|---|---|---|
| Item texts for a quote | Bullet points on work and materials | Every item against your costing |
| Customer e-mail or appointment | Reason, dates, the tone you want | Dates, times, greeting |
| Material list from notes | Your site survey notes | Every measurement against your notepad |
| Daily report for the file | Voice memo or bullet points | Whether it really happened that way |
| Checklist for a routine | How you do it today | Whether a step is missing |
| Small calculator | The calculation rule and one example | The result for a job you know |
All of these have one thing in common: the AI phrases or sorts, you supply the facts and decide. It is more of a writing aid than a skilled worker.
In the trial lesson you put your first instruction together and get feedback. No account, in a few minutes. Start the trial lesson
Quote texts and customer e-mails: you give the facts, the AI the form
Elena is a master painter and writes her quotes in the evening. Her trade software does the costing. What takes time are the item texts: clear for the customer, but precise enough that nobody argues later about whether the ceiling was included. She jots down bullet points from the visit and has ChatGPT phrase the items.
Write item texts for a painting quote to a private customer.
My bullet points:
- living room and hallway, walls and ceilings
- strip old wallpaper, check the surface, fill small cracks
- hang woodchip paper, two coats, color white, matte
- cover floors and furniture, clean up afterwards
Write one item per work step with a short heading and one or two sentences.
Don’t make up quantities, prices or deadlines. I will fill those in myself.
If you are missing a detail, write [missing: …] in its place.
Tone: factual, formal, technical terms only with a short explanation.The key sentence is the one about quantities and prices. Without it, an AI tends to fill in typical values that sound plausible but aren’t yours. With the “[missing: …]” marker you see at once where you still need to add something.
The same goes for customer e-mails. Marco is a car mechanic and has to explain to a customer why the brake discs need replacing as well as the pads. He gives the AI his findings in his own words, his recommendation and the limits: five sentences at most, no workshop jargon, the callback number as a placeholder. The diagnosis is his, not the AI’s.
Check before you send
- Every item and every figure matches your costing.
- Nothing is promised that you don’t want to deliver, such as “including disposal”.
- Dates, times and greeting are right.
- The text sounds like your business. If not, give the AI two old e-mails as a model.
Sorting site survey notes and material lists
Adam is a carpenter. When he measures up for a built-in wardrobe under a sloping roof, he writes measurements on a notepad, and on the drive back he records a voice memo. Turning that into a clean cutting list is tedious work. He types up or dictates the notes and has the AI sort them.
Sort my site survey notes for a built-in wardrobe under a sloping roof into a table.
Columns: part, quantity, length in mm, width in mm, material, remark.
Rules:
- Only use measurements that are in my notes. Don’t convert or round anything.
- If a measurement is missing or unclear, write “?” and list all questions below the table.
- Material: carcass 19 mm white chipboard, back panel 8 mm HDF, fronts 19 mm lacquered MDF.
My notes:
[paste the notes here]The instruction forbids the AI to guess. With measurements that is the most important sentence: a made-up figure in a neat table stands out less than a gap. Adam then puts the table next to his notepad and ticks off every measurement. The questions below the table show him what to measure again at the next visit.
Site reports, checklists and appointments
Erik is a roofer and runs a business with four people. At the end of each day he needs a short report for the file: what was done, what is still open, whether anything held the work up, such as rain or missing material. He used to catch up on this on weekends. Now he records a few sentences on the drive home and gives them to the AI with a fixed structure.
Turn my voice memo into a daily report for the site file.
Structure: date, weather, work done, work still open, hold-ups, material for tomorrow.
Write short, factual sentences. Only include what is in the memo.
If the memo says nothing about a section, write “not stated”.
No names of people, only roles such as “journeyman” or “client”.
Memo:
[paste the memo here]A report like this is only as good as what goes into it. If there is a dispute about delays later, what really happened is what counts. So read it before it goes into the file, and add what the AI could not know.
Checklists
For recurring routines such as servicing a boiler, handing over a site or the final inspection with the customer, the AI can draft a checklist. Victor is a plumbing and heating engineer. He describes to the AI how he works today and asks which steps might be missing. Requirements for the appliance and mandatory test steps he takes from the manufacturer’s documents. The AI sorts; it doesn’t replace a manual.
Appointments
For appointments, a template pays off: a text with placeholders for date, time window and contact person, plus a request to reply by a certain day. Ask for three versions: a short one for a text message, a longer one for an e-mail and one for rescheduling. You fill in the placeholders yourself.
Small calculators the AI builds for you
ChatGPT can also build a small web page that calculates something for you. That is a single HTML file: you save the code with the ending .html and open it in the browser on your computer. Typical examples are calculators for paint quantities, tiles with waste or roof tiles from the roof area. The calculation rule comes from you; the AI puts it into practice.
Build a paint quantity calculator as a single HTML file that runs in the browser without internet.
Inputs: wall area in m², area of windows and doors in m², number of coats, coverage of the paint in m² per liter.
Calculation: wall area minus windows and doors, times coats, divided by coverage.
Result in liters, rounded up to one decimal place.
Below it: how many 10-liter buckets that is, rounded up.
For empty fields or negative values: a clear error message instead of a result.
Large fields and large text, so I can read it on the laptop in the van.Before you trust a calculator, test it on a job you know. Enter the figures for a home whose paint quantity you still remember and compare. Try an edge case too: what happens if windows and doors are larger than the wall? If something is wrong, describe exactly which input gives which wrong result and what would be right.
A good instruction for a calculator like this is like a complete job sheet for the fitter: address, measurements, materials and the exact work are on it, so she can start without calling.
What you don’t hand over
An AI also answers questions it cannot answer reliably, and it sounds just as sure as usual. On these topics, use it at most to sort your questions for an expert:
- Structural calculations and load-bearing capacity: loads, cross-sections, whether a beam will hold. That is calculated by whoever is responsible, such as the structural engineer.
- Standards and regulations: the AI often knows standards only roughly or in an old version. Look up what applies in the standard itself, with your trade association or your chamber of crafts.
- Safety-related values: circuit protection, cable cross-sections, gas and flue gas, tightening torques. What counts here are manufacturer data, tables and your own measurement, not a chat answer.
- Test reports and readings: copy these from your measurement. Don’t let the AI add any values.
- Unchecked prices: hourly rates, material prices and surcharges come from your costing. Typical prices from a chat are not a quote.
- Legal texts: terms and conditions, warranty, acceptance reports. Your chamber, your association or a lawyer can help here.
Customer data: placeholders instead of names
Whatever you type into an AI chat goes to the provider and is stored there. On ChatGPT’s personal plans, such as Free, Go, Plus and Pro, OpenAI also uses your chats by default to train its models. You can turn this off in the settings under “Data controls” with the switch “Improve the model for everyone”. That still doesn’t make a personal account confidential.
In practice this means: names, addresses, phone numbers and photos showing people, house numbers or license plates don’t go into a personal account. Write “[customer]”, “[address]” or “[date]” instead and add the real details afterwards in your e-mail program or your software. The AI doesn’t need them for texts and lists anyway.
If you want to work with real customer data, you need a business plan with a data processing agreement (DPA; in Germany often called AVV). OpenAI offers one for ChatGPT Business and Enterprise. Clarify with whoever handles data protection for you whether that suits your business. How the services differ is explained in the guide Which AI is right for me?.
How to practice this
The core of every example is the same: say exactly what should come out, for whom and by which rules, and check the result against something you know. That can be practiced. In the app “KI können” you write instructions for tasks from your own job in short lessons and put them to work in your own AI. The trial lesson takes a few minutes and needs no account. You can see every topic in the course overview.
Sources
- OpenAI Help Center, Data Controls FAQ (opens in a new tab)
- OpenAI, Enterprise privacy at OpenAI (opens in a new tab)
- General Data Protection Regulation, Art. 28 Processor (EUR-Lex) (opens in a new tab)
Information as of November 15, 2026. Laws and offers can change; this article is not legal advice.
Read more
Try it now
The trial lesson with examples for Cabinetmaker. No account, a few minutes.
Start the trial lesson