AI for plumbing firms: booking more work without answering more calls
A plumber under a sink cannot answer the phone. The customer ringing does not know that, does not care, and mostly does not leave a voicemail: they ring the next firm on the list, and the missed call quietly becomes someone else's booked job. The way out is not answering more calls. It is making sure every enquiry gets an instant acknowledgement and a useful next step without you touching the phone.
That is now a solvable problem with tools priced for a two-van firm rather than a plc. The pieces are enquiry capture, diary booking, quote drafting from photos, and follow-up. None of them require anything resembling programming, and none of them are magic: each one takes a decision from you once, then applies it every single time.
This post sets the stack out in the order it pays back, and is honest about where automation must stop: emergency triage still needs a human, fast.
Why do plumbing firms lose work they never see?
Because the lost enquiry leaves no trace. The job that booked with a competitor never appears in your diary, your accounts or your job sheets. It shows up, if at all, as a missed call in a phone log nobody reviews, so the cost never gets counted, and uncounted costs never get fixed.
The arithmetic is worth doing openly rather than quoting invented industry percentages. Call it three missed calls in a working week, and suppose just one of the three would have become a job at your average invoice value. Whatever that value is for your firm, multiply it by fifty weeks and you have the rough annual price of being under a sink when the phone rings. For most firms it pays for a great deal of automation.
What does a practical AI stack look like for a plumbing firm?
Four layers, in payback order. Each works on its own; together they behave like an office manager who never sleeps.
| The job | What runs without you | What lands back with you |
|---|---|---|
| Enquiry capture | Every call, WhatsApp and web form gets an instant reply asking the useful questions: where is the leak, can you send a photo, is the water off | A structured enquiry with photos attached, not a voicemail |
| Diary booking | Non-urgent work is offered real slots from your calendar; confirmations and reminders go out on their own | A booked job with address, photos and access notes |
| Quote drafting | A draft assembled from the customer's photos and your own price list | A document to check and send the same evening |
| Follow-up | Unanswered quotes get a polite nudge at day three and day ten, then stop | Replies threaded back to the job, plus a short list worth a phone call |
The common thread is that nothing in that table waits to be prompted. A chatbot you must remember to use is another job on the list. This is the opposite: the work arrives partly done.
Can AI really draft a quote from photos?
Within limits, yes, and the limits matter. A clear photo of a leaking trap, a boiler nameplate or the pipework under a sink is enough for a model to identify the fitting, draft the description and pull a provisional price from your price list (yours, not one it invents). What it cannot see is corrosion inside a wall, a seized isolation valve, or the reason the last plumber gave up. So the working rule is simple: the system drafts, you price. Thirty seconds of checking replaces forty minutes of evening typing, and the customer has a written quote while your competitor is still promising to get back to them.
Two boundaries belong in the wording of any automated quote. Anything involving gas is carried out by a Gas Safe registered engineer, which is a legal requirement rather than a preference, and boiler work should never be priced as final from photographs alone. And if you hold approved-contractor status under a scheme such as WaterSafe, the automated messages should claim exactly what the scheme says and nothing grander.
Where must a human still take over immediately?
Emergency triage. A caller reporting a smell of gas is not a booking conversation, and an assistant that offers them Thursday at ten is a liability. The system should be built to recognise those words, tell the caller to ring the National Gas Emergency Service straight away, and stop. Water pouring through a ceiling deserves one automated message (where the stopcock usually is, send the address and a photo) and then a human voice. The honest design principle: automation buys minutes and captures facts, but it never decides whether to attend.
Say what it is, too. "You are talking to our booking assistant, a human will confirm shortly" reads fine at 9pm. Pretending the machine is a receptionist reads terribly the moment it misunderstands something.
What about the customer data you are now capturing?
Names, addresses, phone numbers and photographs of the inside of people's homes are personal data, and UK GDPR applies to a two-van firm exactly as it applies to a plc. The ICO's guidance for organisations is the reference worth an hour of your time. In practice: know which tool holds your enquiry data, do not let job photos live forever in a personal WhatsApp, delete what you no longer need, and be able to answer "where is my data?" in one sentence. Tools with a UK or EU storage option make that sentence shorter.
The plumbing firms winning locally are not answering more calls. They have made the first five minutes of every enquiry automatic, so the human conversation starts with the job already half-captured. If you would rather have this built than assemble it from a dozen subscriptions, workflow automation is the kind of work AI Metric does, but the stack above is achievable either way. The phone will still ring while you are under a sink. The difference is what happens next.