How to Use AI in a Plumbing Business: A 30-Day Setup Plan

Spend week one on calls, week two on quotes, week three on follow-up and week four on review, tracking one number each week. Add one tool a week, keep emergencies and prices in human hands, and cut anything that has not moved its number by day 30.

The order matters because each week feeds the next. A missed call becomes a quote request, a quote becomes a follow-up, and a follow-up becomes a review request. You need about three hours a week from the owner or office manager, plus a notebook or spreadsheet for the numbers. Most firms go wrong by buying a platform on day one and switching everything on at once, which makes it impossible to tell what helped.

Follow me on Instagram@sagnikteaches

Four weeks, four numbers: the plan on one sheet

Print this table and stick it where the phone lives. The tools named in the sections below are examples; the method works with whichever job software you already use.

Connect on LinkedInSagnik Bhattacharya

If you run a mixed trades business rather than a plumbing firm, the trade-neutral version of this plan is the first 30 days of AI in a trades business. This one is written around plumbing calls: leaks, blockages, no hot water and the occasional emergency.

Subscribe on YouTube@sagnikteaches

Before day one: a 45-minute baseline

You cannot tell whether a tool helped without a starting figure, and nobody remembers their own numbers accurately. Spend 45 minutes pulling four things from the last 30 days: calls received and calls missed (your phone's call log or your provider's app), the time to first reply on 10 missed calls, the hours between visit and quote on your last 10 quotes, and how many of the quotes you sent were chased and how many were accepted. Then add an honest guess at your weekly admin hours.

Here is a baseline for an illustrative three-van firm with a part-time office manager. Your figures will differ; what matters is that you write them down.

The rest of the plan reports against this table. The 22% miss rate here comes from a firm whose plumbers are under sinks all day, and each missed call may be a customer ringing the next name on the list.

Week one: calls

Calls come first because a missed call is the one loss you cannot recover afterwards: the customer has already rung someone else. The week has four jobs, one for each of the first four working days, with the go-live on day five.

Write the phone rules before you buy anything

Spend 30 to 45 minutes writing down how every kind of call should be handled. The same rules brief an office manager, an answering service or an AI tool, and writing them is where most owners discover they have never decided what counts as an emergency. A starting version for the AI receptionist's instructions:

You answer calls for [business name], a plumbing and heating firm. Hours: [hours]. We cover [service area in one sentence].

Always say you are an AI assistant. Take the caller's name, phone number, address and a one-line description of the problem, then say a person will call back.

Emergencies (alert the on-call plumber on [number] straight away, do not wait for the callback): water that cannot be isolated, flooding, sewage coming back into the property.

Gas smell, or a carbon monoxide alarm sounding: tell the caller to leave the property, not to touch switches or flames, and to call the gas emergency service or the emergency number. Do not book a visit for these calls.

Never: promise an arrival time, quote any price except the call-out fee of [fee], diagnose a fault, or give advice about gas work.

If you do not know, say a plumber will call back. Do not guess.

Notice what the block refuses to do. It does not diagnose, price or promise, because those are the three things a confident-sounding assistant will do if nobody forbids them. Tell callers they are speaking to an AI assistant at the start; it is honest, it reduces the sense of being tricked, and some places require it. The firm's emergency list belongs to you, not the tool: if you would send a van out at 2am for it, it goes in the emergency line. For a deeper treatment of burst-pipe rules and booking limits, see AI receptionists for plumbers: cost, burst-pipe rules and booking limits.

Choose the lightest tool that covers your gap

There are three routes, in order of weight. Start with the lightest one that closes the gap you measured.

• Missed-call text-back and voicemail transcription. The cheapest fix: an instant text saying you are on a job and will call back within the hour, plus a written transcript of the voicemail. Many business phone systems include it.
• An AI receptionist inside your job software. Jobber lists its Receptionist at $29 a month as an add-on on select plans, and its help article says that covers 30 conversations, then $0.79 each (as checked in late October 2026). Housecall Pro and ServiceM8 sell their own versions, and their pricing pages do not always say which plan includes what, so ask before you assume.
• A standalone answering service. More control, usually more cost. A live-staffed plan on one provider's pricing page starts at $300 a month for 30 calls, which shows how far apart human and AI answering sit on price.

A quick sum for the illustrative firm. If all 41 missed calls became receptionist conversations, the bill would be $29 plus 11 conversations over the allowance at $0.79, which is $29 + $8.69 = $37.69 a month. Many missed calls are hang-ups or sales calls, so treat that as a ceiling.

Test calls that expose the weak spots

On day three or four, ring the line from another phone and play the awkward customer. Run each of these at least once, and write down exactly what the assistant said.

Expect to fail two or three on the first run. A typical failure reads like this: the caller says there is water coming through the kitchen ceiling and the assistant answers "I can have someone with you within 30 minutes." That is a promise the firm cannot keep, so "never promise an arrival time" goes into the rules and the test is run again. Fifteen calls is a better test than six; the 15-call test for plumbers gives the full list.

Go live on after-hours and overflow first

On day five, route only the calls you would otherwise miss: after hours, and calls unanswered after a set number of rings. Keep daytime answering human until the transcripts look right. For the first week, read every transcript each evening; it takes about ten minutes and it is the cheapest quality control you will ever buy. Add a line for each mistake to the rules and retest the call that caused it.

The week-one number is the median minutes from a missed call to the first reply, whether that reply is from the assistant or a person. In the illustration it falls from 3 hours 40 minutes to 22 minutes, because the assistant answers within a minute and the human call-back happens inside the hour.

Friday checkpoint (20 minutes). Three questions. Did any emergency go down the wrong path? If yes, pause the tool until the rule is fixed. How many transcripts needed a human correction? And what is this week's median reply time against the baseline?

Week two: quotes

Quotes are the second loss. A quote that lands two days after the visit often lands after the customer has picked someone else. Week two has the plumber speak into a phone and the assistant turn the words into a draft, with your price book doing all of the pricing.

Put your prices in one place

On Monday, write the price book the assistant will be forced to use. It needs the call-out fee and what it covers, an hourly rate, a minimum charge, an out-of-hours multiplier, your parts markup rule, how long a quote stays valid, and ideally 20 to 30 flat-rate jobs. The flat-rate list is the long part, and a worked 30-task plumbing price list shows how to build it. The extract below is illustrative, not a recommendation for what to charge.

From voice note to draft quote

On Tuesday and Wednesday, the plumber records a 60 to 90 second voice note at the end of each visit, always in the same order: who the customer is, what you found, what you will do, what you will not do, how long, parts with costs, and when the customer is in. Dictate it with the phone's own dictation or an assistant app's voice mode, then paste the text. An illustrative note for a small bathroom job:

Mrs [surname], upstairs bathroom. Basin has old pillar taps, she wants a single mixer, I'll supply it. Cistern keeps running, fill valve has gone, replace it. No isolation valves on either, so fit three, fifteen mil. Two hours on site. Parts cost: mixer seventy, fill valve twenty-five, three isolators eight each. She's in Tuesday or Thursday mornings.

The prompt that turns that into a quote:

Turn this voice-note transcript into a quote for the customer. Use my price book below for every price. If an item is not in the price book, write [price needed] and do not guess.
Sections: what we found, what we will do, what is included, what is not included, price (labour, parts, total), how long it will take, how long the quote is valid.
Do not add guarantees, discounts or promises that are not in my notes. Plain English, short sentences.

Price book: [paste]
Voice note: [paste]

A realistic first draft (illustrative) gets the structure right and the numbers slightly wrong. It prices labour at "2 hours at $95 = $190", because it has used the call-out fee as the hourly rate, and it adds "all work carries a 12-month workmanship guarantee", which the firm has never offered. The plumber, who is the only person who knows either fact, makes two corrections.

The $20 gap is small, and that is the danger. Errors like this survive because nothing looks wrong. If your job software can draft quotes from a voice note itself, test it on the same notes and compare; Jobber's AI Voice, for instance, lists creating quotes among its tasks, though its pricing page does not say which plans include it. For more on whether quote writing is safe to hand over at all, read can a plumber use AI to write quotes.

Decide which quotes need the owner's eyes

On Thursday, set a rule so review does not depend on how busy you are. The person who visited sends every quote, after reading it against the price book. The owner also reads any quote above a threshold you choose (say $1,000), anything involving a heating appliance, gas or drainage work, and any customer type you have not quoted before. Nothing goes straight from the assistant to the customer.

The week-two number is the median hours from site visit to quote sent. In the illustration it drops from 26 hours to 5, because quotes are drafted the same afternoon from the voice note.

Friday checkpoint (20 minutes). Did any quote go out with a price that did not come from the price book? Did any draft add wording you do not offer, such as a guarantee, a discount or a timescale? And what is this week's median against 26 hours?

Week three: follow-up

Most quotes that go quiet are not refusals. The customer is busy, or comparing, or waiting for a nudge, and a short specific message wins some of them back. Week three covers three follow-ups: quote chasers, review requests and invoice reminders. Do the quote chasers first, because they are closest to money.

A three-touch quote chaser

Use the same three touches for every quote. On day two, check it arrived and invite a question. On day five, offer a specific slot. On day twelve, say the quote lapses on its expiry date and that you will close the file unless you hear otherwise. Stop after the third touch, stop the moment someone says no, and never chase a customer who has asked you not to. Some job software schedules reminders for you, so check your plan before building it by hand. If not, a weekly task works: list every open quote older than two days and let the assistant draft the next message for each.

Here are my open quotes: customer first name, job, date sent, amount.
For each one, draft the next message using this rule: touch 1 (day 2) checks the quote arrived and invites a question; touch 2 (day 5) offers a specific slot next week; touch 3 (day 12) says the quote lapses on [expiry date] and I will close the file unless I hear otherwise.
Keep each message under 60 words. Friendly and plain. No discounts, no pressure, no promises I have not listed. Use [first name] and [your name] as placeholders. Return a table: first name, touch number, message.

[paste the list]

Before and after, for touch two on the bathroom quote:

• Before: "Hi just checking if you got our quote."
• After: "Hi [first name], it's [your name] from [business]. I sent your bathroom tap and cistern quote on Tuesday. The price holds until the 28th, and I can do it on a morning visit next week. Shall I pencil one in, or is there anything in the quote you would like me to change?"

The second version asks one question and offers one action. Someone reads it, checks the customer has not already booked, and sends it. The most common error here is not the wording but the data: a chaser sent to a customer who accepted yesterday. Mark accepted quotes before the weekly list is built.

A quick sum shows why this is worth an afternoon. Suppose the 38 quotes a month convert at 24%, and chasing lifts that to 29%. That is five percentage points, or about 1.9 extra jobs a month. At an average job value of $480, that is roughly $912. One month of data cannot prove the cause, so keep the log for three months before claiming it.

Review requests: ask everyone, the same day

Send a short text the same afternoon or next morning, with your review link, to every customer. Not only the happy ones. Google's policy for contributions to its maps lists "selectively solicit positive reviews from customers" and offering incentives such as discounts in exchange for reviews as prohibited behaviour. So the rule is simple: ask everyone, and offer nothing in return.

• "Hi [first name], thanks for having us in today. If you have a minute, an honest Google review of your experience helps other customers decide. Here is the link: [link]. If anything was not right, reply here and I will sort it personally."

The message asks for an honest review and does not suggest what to write, because Google's policy also says merchants should not request that specific content be included. The last sentence is not a filter; the link goes to everyone. It gives an unhappy customer a faster route to you than a one-star review. For the wider routine, see getting more reviews after every job with AI follow-ups.

Invoice reminders in the same sitting

While the follow-up list is open, add the invoice reminders: a friendly note a few days after the due date, and a firmer one a week later, both from the same template. Keep the template in your assistant's saved instructions so the tone stays yours.

The week-three number is the share of quotes chased within three days of sending. In the illustration it moves from 29% to 89%: 8 of the 9 quotes sent that week were chased on time.

Friday checkpoint (20 minutes). Did anyone complain about a message? Did a chaser reach a customer who had already booked? And how many chasers got a reply, even a "no thanks"? A "no thanks" is a result.

Week four: review and decide

On Monday, rebuild the baseline table with this month's figures. Then work out what the tools cost, and decide for each one: keep, hold or cut.

The monthly cost in this illustration is $29 for the receptionist add-on plus about $20 for an assistant plan, so $49. Divide that by your own hourly cost to get the break-even: at $25 an hour it is about two hours a month, and the admin figure fell by four hours a week. That sum favours keeping both, but the table, not the sum, is what you act on. Use a plain decision rule:

Finish the month by writing a five-line note for the team, so the rules outlive your memory of them:

How we use AI here
1. The phone assistant answers after hours and overflow. Emergencies go to the on-call plumber, never to the assistant.
2. Quotes are drafted from voice notes and priced from the price book. Whoever did the visit checks and sends them.
3. Follow-up messages go out on days 2, 5 and 12 and stop at "no".
4. Customer names, addresses and numbers go only into tools on our business plans.
5. Anything the assistant gets wrong goes in the mistakes log, which we read every Friday.

What to skip in the first 30 days

Several good ideas do not belong in month one, because they add a second thing to manage before the first is under control.

• A website chatbot. It creates another inbox. Get calls and quotes stable first, then add it.
• A messaging platform. Podium's plan cards list an entry price of $399 a month, several times the whole plan above, before you know which part of it you would use.
• AI-written social posts on a calendar. They do not fix calls, quotes or follow-up, which is where jobs leak.
• AI scheduling and route optimisation. It needs clean job data and a steady schedule, which month one is building.
• Google Ads automation. Spending more to attract calls you are already missing is the wrong order.
• Anything that sets prices. The price book sets prices. The assistant formats them.
• Autonomous agents that act inside your accounts. Keep a person between the draft and the send until the mistakes log has been quiet for a month.

After day 30, the next additions follow your numbers: a maintenance-plan offer to past customers if follow-up worked, a website chatbot if enquiries arrive out of hours, or a closer look at job software if the spreadsheet is straining. If you would rather map the sequence with someone, my AI implementation consultation covers exactly that.

Questions about the 30-day plan

Can a one-van plumber follow this plan?

Yes, with less. Skip the AI receptionist at first and use a missed-call text-back plus voicemail transcription in week one. Keep weeks two to four as written, because quotes and follow-up are where a one-person firm loses the most time. Expect about two hours a week of your own time instead of three.

Do I need job management software before starting?

No. Week one needs only a phone and a call log, and weeks two and three work with a spreadsheet and a general AI assistant. If you already run job software, check it first, because many now include a receptionist, quote drafting or reminders, and using those means fewer logins and one customer record.

What if the AI receptionist books or promises the wrong thing?

That is what the test calls and the first-week transcript reading are for. Read every transcript for the first seven days, add a rule for each mistake, and test again. Keep emergencies on a human path from day one, and never let the tool promise arrival times or prices it cannot know.

What should the first month cost?

Between nothing and about $50 a month for a small firm. An AI receptionist add-on can start around $29 a month, and a general assistant plan is about $20 a month. The real cost is your three hours a week, so measure the numbers before deciding what to keep.

Further reads

• How Can AI Help Plumbers? Start With Calls, Quotes or Follow-Up — Start with calls, quotes or follow-up: which first?
• 10 ChatGPT Prompts for Plumbers: Call-Backs, Quotes, Complaints — Ten ready-to-paste prompts for call-backs and quotes.
• Jobber AI Receptionist Pricing: Is It Worth It for a One-Van Plumber? — What one receptionist add-on costs a one-van firm.
• Best AI for Plumbers: What a Solo, Two-Van and Ten-Van Firm Should Buy — What a solo, two-van and ten-van firm should buy.
• Plumbing Maintenance Plans: AI-Written Offers and Renewals — A good use of week-three follow-up time.
• How to Set Up Missed Call Text Back for a Service Business — Set up the lightest week-one tool.
• Chasing Unpaid Invoices With AI: Polite Reminders for Trades — Polite reminders for unpaid plumbing invoices.
• AI Automation for Plumbers: Package Prices and What Should Be Included — Done-for-you AI packages for plumbers priced from the bottom up: inclusions, a 12-month sum against doing it yourself, ownership and contract checks.
• AI Software for Plumbers: What a Small Firm Pays Each Month, by Van Count — Monthly AI software bills for one, three and eight vans, tallied line by line from vendor pricing pages: platform, AI receptionist and ChatGPT seats.
• Plumbing Lead Generation: Cost per Job Tracked With AI Summaries — A monthly tracker for plumbing lead sources: tag calls with AI summaries, divide spend by paid jobs, and keep, fix or cut each source.
• AI Chatbot for Plumbers: Answer Questions, Collect Photos, Refuse Jobs — Set the scope of a plumber's website chatbot: three lanes, an approved answer sheet, price rules, photo requests, refusals, handover and a 20-message test.
• AI Marketing for Plumbers: A Worked 12-Month Posts and Offers Calendar — A worked 12-month calendar for a five-van plumbing firm: frozen pipes, boiler checks, heater age and drain season, with AI drafts and a review step.
• Plumbing Van Stock Lists: Set Reorder Levels From Job History With AI — Export a year of parts, merge duplicate names with AI, rank the top 40, set minimum and maximum levels per van and check them against supplier stock.
• Best AI for Plumbing Questions: A Ten-Question Test for Plumbers — A ten-question test plumbers can run on ChatGPT, Claude and Gemini, with a scoring sheet and the checks to make before acting on any answer.
• Can AI Tools Help With Plumbing Design? What Needs a Qualified Person — What AI can draft in a plumbing layout, what a licensed designer must verify, and a sign-off checklist for the contractor.
• Can AI Turn Drain Camera Notes Into Customer Reports? — A worked notes-to-report example for drain camera surveys, with a prompt, a sample report and the engineer's sign-off checks.
• Can AI Update Your Plumbing Price Book From Supplier Invoices? — Update a plumbing price book from a wholesaler invoice with AI: line extraction, cost comparison, markup rules and the checks that catch mistakes.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Jobber's pricing page and Receptionist help article, Google's Maps contributor policy, and the Housecall Pro and ServiceM8 help and pricing pages, read in October and November 2026. The three-van firm is an illustration, not a client result.