AI Workflows for Plumbers and HVAC Techs: Real Uses, No Hype

Forget everything the AI companies say about revolutionizing the trades. No chatbot is snaking a drain or brazing a line set. But there is a quieter, real version of AI in service businesses, and it lives in the office work around the job: the notes, the messages, the follow-ups, the reviews. That work eats an hour a day, and AI does it well, right now, with tools that cost nothing.

Three workflows. Each one takes fifteen minutes to set up and saves that much every week. No hype attached.

Workflow 1: dispatch notes to clean tickets

The call comes in: "Water heater's making a noise, kind of a rumbling, and there's water on the floor but not much, it's in the garage." The dispatcher types what they hear, which is a paragraph of fragments. The tech reads it in the driveway and calls back for clarification. Twice.

The workflow: paste the raw call notes into the AI with a prompt that says "turn this into a dispatch ticket: customer issue in one line, suspected cause, what to bring, questions to ask on arrival." Out comes a clean ticket: "Rumbling water heater with minor floor water, garage. Suspected sediment buildup or failing T&P valve. Bring: element kit, T&P valve, wet vac. Ask: age of unit, any recent work."

The dispatcher still takes the call. The tech still diagnoses on site. The AI just converts fragments into structure in ten seconds. Multiply by fifteen calls a day and the dispatcher gets an hour back. The tickets are also more consistent, which means fewer "what did they say the problem was" callbacks from the field.

The prompt for this lives wherever the dispatcher works: a printed card by the phone, a note on the computer, a text file. Copy, paste the notes, paste the result into the ticket. The whole workflow is thirty seconds per call.

Workflow 2: quote follow-ups that get answered

The tech quotes a $2,400 repair on site. The customer says "let me think about it." The quote goes into the system, and then nothing happens, because the tech is on the next call and the office is dispatching. Three days later the customer has either called someone else or forgotten.

The workflow: every open quote over 48 hours old gets a follow-up draft from the AI. The prompt takes the quote details (customer, job, amount, date) and drafts a short message: reference the specific job, one useful detail from the visit, one question. The office reviews it, which takes twenty seconds, and sends it by text.

The key is the detail from the visit. "Following up on the water heater quote" is ignorable. "Following up on the water heater quote, the 12-year-old unit in the garage. Did you want me to hold Thursday for the install?" is answerable. The prompt is instructed to include one specific from the ticket notes, and the reviewer checks that it did. Specific beats generic every time, and the AI is perfectly capable of specific when the notes are.

Shops that run this workflow consistently report the same thing: a meaningful share of "thinking about it" quotes come back. Not because the message is clever, but because it exists. Most competitors never follow up at all.

Workflow 3: review replies in your voice

Reviews are the trade's word of mouth, and replying to them matters: it shows future customers that the shop is alive and accountable. But writing a thoughtful reply to every review is exactly the kind of task that slides to "later" permanently.

The workflow: paste the review into the AI with a prompt that says "draft a reply in the voice of a straightforward local service company: thank them, reference the specific work mentioned, invite future business. Keep it under 60 words." The draft comes back, the owner or office manager reads it, adjusts anything that sounds off, and posts.

For negative reviews, the prompt changes: "draft a public reply that acknowledges the issue without admitting fault, states the specific next step, and moves the conversation to a phone call. Calm, professional, no defensiveness." The human reviews these with extra care, because a public reply is permanent. But starting from a calm draft beats starting from a defensive reaction at 10 p.m., which is when most review replies get written.

The voice instruction matters. Generic AI replies ("Thank you for your feedback! We strive for excellence!") are worse than no reply, because customers can smell them. The prompt should include two or three phrases the shop actually uses. Feed the AI your voice and it will write in it. Feed it nothing and it writes like a hotel chain.

The no-hype rules

Three workflows, three rules that keep them safe:

Read everything before it goes out. Every draft, every time. The AI does not know the customer, the job history, or what was promised on site. The thirty-second read is the entire quality control department. No exceptions, including for "just a review reply."

AI drafts words. Humans make commitments. The AI never sets prices, never promises arrival times, never authorizes warranty decisions. It turns notes into tickets and tickets into messages. The moment a message contains a commitment, a human wrote that part.

Keep customer data sensible. First names and job details are fine. Do not paste anything you would not want on a postcard: no full addresses in prompts you do not need to, no payment details ever. The workflows above do not need sensitive data, so do not give it.

These rules are not anti-AI caution. They are what make the workflows trustworthy enough to run every day. A workflow the owner has to worry about does not get used. A workflow with clear guardrails becomes boring, and boring is the goal. Boring means it runs.

Start with one

Do not implement all three on Monday. Pick the one that hurts most. If dispatch notes are chaos, start there. If quotes die in silence, start with follow-ups. If the Google profile has twelve unanswered reviews, start there. Run it for two weeks until it is boring, then add the next. File the prompts alongside the other systems and tools the shop runs on: the estimate template, the ticket log, the price book. They belong together.

That is the honest version of AI in the trades. Not a revolution. Three small workflows that convert fragments into structure, silence into follow-up, and reviews into replies. An hour a day back, for the price of copy and paste.

This is the built-with-intelligence idea in its simplest form: AI that plugs into the work you already do, through the AI service you already use, with prompts written for your trade instead of generic business advice. No new subscription, no new software, no per-seat anything. Just better words, faster.

The Contractor Prompt Pack is 50 copy-paste prompts in this spirit, including the dispatch-note cleanup, the quote follow-up drafts, and the review reply prompts above: each with a title, when to use it, the prompt, and a usage tip, in a formatted playbook PDF plus a plain-text library for copy-paste from your phone. $29 as a one-time purchase, no subscription, and the prompts work with the free tier of ChatGPT, Claude, or Gemini.