Key takeaways
- Use AI for wording, structure and clarity; keep scope, quantities, price, code and warranty decisions with a person.
- Preparing cost estimates is a core plumber duty[1], and a lead tech's time is expensive; the median plumber earned $30.67 an hour in 2025[2].
- Feed the tool facts from the tech's notes and photos, then ask it to rewrite, never to fill gaps.
- Ban unsourced claims about savings, code requirements and lifespans in AI drafts.
What AI should and should not do in an estimate
An estimate does two jobs. It sets the price and scope, which is a business and technical decision, and it explains the work in a way the customer understands, which is a writing task. AI is useful for the second job and risky for the first. Keeping that line clear is the whole method.
Let AI do these things:
- Turn a tech's shorthand notes into clear sentences a homeowner can follow.
- Structure the description: what was found, what will be done, what is included, what is not.
- Write two or three option descriptions from the scope you already decided.
- Adjust reading level and length, for example a short version for a text and a fuller one for the PDF.
- Draft a translation for a fluent speaker to check.
Keep these with people:
- Scope, quantities, materials and labor hours.
- Price, discounts, financing terms and deposit rules.
- Code requirements, permits and inspection steps.
- Warranty terms, exclusions and any statement about how long something will last.
- Anything about the cause of a failure or who is responsible for it.
Why it is worth doing
The Bureau of Labor Statistics lists preparing cost estimates for clients among the core duties of plumbers[1]. In many shops the person writing estimates is a lead tech or service manager, sitting in the truck at the end of a long day. Their time is valuable: hourly wages for US plumbers ranged from $21.22 at the 10th percentile to $52.13 at the 90th in 2025, with a median of $30.67[2], before benefits and overhead.
Clear estimates also sell better. A homeowner who understands what will be done, and why, is more likely to approve the work and less likely to dispute the invoice. AI does not make a tech a better plumber, but it can make a good plumber's explanation easier to read. Use is still modest in the trades: 19.8% of construction businesses said employees used generative AI for work tasks in a recent Census survey[3], so shops that do this well can stand out.
The workflow: notes in, draft out, sign-off
Capture facts on site
The tech records what was found, measurements, model numbers, access issues and the agreed options in the job notes, with photos. These are the only facts the draft may use.
Remove personal details
Before pasting into any AI tool, remove the customer's name, address and phone number. Use the job number instead.
Generate the draft
Paste the notes with a standard prompt from your prompt library, asking for a rewrite, not new content.
Insert price and terms yourself
Prices, quantities and warranty wording come from your price book and templates, never from the draft.
Check against the notes
Read the draft line by line against the tech's notes. Delete anything the notes do not support.
Approve and send
The service manager or estimator approves the final version, and their name is on it.
Prompts that work
Build a small library of prompts and keep using the same ones; consistency matters more than clever wording. These are good starting points:
- Rewrite these technician notes as a clear job description for a homeowner. Use only the facts given. Do not add causes, savings, code requirements or timeframes. Sections: what we found, recommended work, what's included, not included.
- Write three short option descriptions from this scope: Option 1 repair the existing valve; Option 2 replace the valve; Option 3 replace valve and supply lines. Keep each under 60 words. Do not mention prices.
- Shorten this estimate description to four sentences for a text message, keeping the work described exactly the same.
- List any statements in this draft that are not supported by the notes below.
- Rewrite at an eighth-grade reading level without changing meaning.
The fourth prompt is the most underused. Asking the tool to check its own draft against the notes catches many invented details before a person even starts reading, though it does not replace that read.
Review rules for every AI draft
| Element | Rule | Approved by |
|---|---|---|
| Scope and quantities | Must match tech notes exactly | Service manager |
| Price and terms | From price book and templates only | Estimator |
| Code and permits | Only statements the office has confirmed for that jurisdiction | Licensed plumber |
| Savings or efficiency claims | Only with a named source, or removed | Service manager |
| Warranty | Standard warranty text, never AI wording | Owner |
| Cause of failure | Factual observation only, no blame | Lead tech |
Watch the claims and the numbers
AI tools like to add persuasive extras: this will slash your water bill, this meets all current codes, this will last for decades. Each of those is a claim you would have to stand behind. Remove any claim that does not have a source you can name, and where a sourced fact helps the customer, use the source's own wording.
For example, if an estimate covers repairing several small leaks, the EPA states that fixing easily corrected household leaks can save homeowners about 10% on their water bills[4]. That is a reasonable line to include, attributed to the EPA. An AI-generated figure with no source is not.
Code statements need even more care. Requirements vary by state and city, and inspectors have the final word. Write what you will do, such as obtain the permit and schedule the inspection, rather than general claims about what code requires.
Example: before and after
Example tech notes: 40 gal gas WH, 14 yrs, tank seeping at base, rust on burner floor, no exp tank, TPR discharge line short. Options 1 like-for-like 40 gas, 2 same plus exp tank, 3 tankless, customer wants 1 or 2.
Example AI draft after review: What we found: your 40-gallon gas water heater is about 14 years old and the tank is seeping water at the base, with rust under the burner. Tank leaks cannot be repaired, so we recommend replacement. Option 1: replace with a new 40-gallon gas water heater, including removal of the old unit. Option 2: the same as Option 1, plus a thermal expansion tank and extending the relief valve discharge line. Not included: changes to gas piping beyond the connection, or drywall repair. The office added prices from the price book and the standard warranty text, and removed one sentence the draft had added about energy savings.
Job descriptions after the work is done
The same method works for the other half of the paperwork: the description of completed work on the invoice and in the job history. These records matter later, when a warranty claim arrives, a customer disputes a charge or a different tech returns to the same house two years on. Vague entries such as replaced valve, all good help nobody.
Ask the tool to turn the tech's closing notes into a short, factual record with a fixed structure: what was found, what was done, parts installed with model numbers, tests performed, and anything recommended but declined by the customer. That last line is the one most often missing and the one that protects you most when a declined repair fails later. Keep the record factual and neutral, without sales language, and have the tech confirm it before the job is closed.
Over time these structured records also make your reports more useful. When every job history uses the same headings, it becomes much easier to search for repeat failures, check which parts are being used, and spot techs whose notes need coaching.
Rolling it out to the team
- Start with one estimator and ten estimates, comparing AI-assisted drafts with their usual ones.
- Agree the prompt library and the review table before anyone else starts.
- Teach techs to write better notes; the quality of the draft depends on them.
- Track approval rates and customer questions on AI-assisted estimates against your usual rate.
- Revisit the rules after a month and tighten any area where drafts needed heavy correction.
Our guide on good-better-best options covers how to structure the choices the AI helps you describe.
Frequently asked questions
Can AI write a plumbing estimate for me?
It can draft the wording from your tech's notes. Scope, quantities, price, code statements and warranty terms should always come from your own people, price book and templates.
Is it safe to paste job details into an AI tool?
Remove customer names, addresses and phone numbers first and use a job number. Use a business account with training on your inputs switched off where possible.
How do I stop AI adding claims I can't back up?
Tell it not to add causes, savings, code requirements or timeframes, ask it to list unsupported statements, and have a person delete anything the notes do not support.
Will customers notice the estimate was AI-assisted?
They should notice it is clearer. What matters is that every fact is right and a named person at your company stands behind it.
References
U.S. Bureau of Labor Statistics (2025). Occupational Outlook Handbook: Plumbers, Pipefitters, and Steamfitters (opens in a new tab) โ
O*NET OnLine (U.S. Department of Labor), BLS 2025 wage data (2025). National Wages 47-2152.00 - Plumbers, Pipefitters, and Steamfitters (opens in a new tab) โ
U.S. Census Bureau (2026). Business Trends and Outlook Survey (BTOS), response estimates by sector (opens in a new tab) โ
U.S. Environmental Protection Agency, WaterSense (2026). Fix a Leak Week (opens in a new tab) โ