Key takeaways
- Automate predictable, factual messages: booking confirmations, reminders, 'on my way', job complete and invoice sent.
- Deliver bad news in person: long delays, price changes, unsafe appliances, failed repairs and complaints.
- Construction lags on automation and AI: 81% of construction businesses used no AI in autumn 2026, against 64.1% of all businesses[1].
- Keep service messages free of promotions; marketing texts and emails need consent unless the soft opt-in applies[2].
- A person should check templates and exceptions; most businesses using AI already check its outputs[3].
Why automate the routine messages
Most of the messages a plumbing or heating firm sends are predictable. Every booked job needs a confirmation, a reminder the day before, a warning when the engineer sets off and a note when the invoice goes out. Sent by hand, those messages depend on someone remembering at the end of a long day, and they are the first thing to slip when the phones are busy. Sent automatically from the job system, they go out every time, at the right time, in the same words.
That consistency is the main benefit. Customers who know when the engineer is coming, and what happens next, ring the office less and complain less. The risk is automating the wrong things: news that needs a human voice, delivered by a template that sounds like it does not care. This guide draws the line.
Where the trades are with automation
If you have not automated much yet, you are in the majority. The ONS business survey in autumn 2026 found that most construction businesses were not using any AI technology, a higher share than across the economy as a whole[1].
Source: Office for National Statistics, 2026[1]
Show the data
| Item | Value |
|---|---|
| Construction businesses | 81% |
| All businesses | 64.1% |
| Businesses with 0 to 9 employees | 65.7% |
Among the construction businesses that did use AI, 65.1% used it to improve business operations and 15.6% to provide or personalise services for customers[1]. Customer updates sit between the two: they are an operations job that customers feel directly. You do not need AI to send a reminder, though; most job management systems can send confirmations and reminders from simple rules, and that is the right place to start.
What to automate
Automate a message when the content is factual, the trigger is clear and the customer would be annoyed not to get it. These are the usual candidates:
| Message | Trigger | What it must say |
|---|---|---|
| Booking confirmation | Job booked | Date, arrival window, what the visit covers, call-out charge basis, how to change it |
| Reminder | Day before the visit | Window, engineer's name, parking or access request, how to rearrange |
| On my way | Engineer marks the previous job complete | Engineer's name, rough arrival time, a number to call |
| Job complete | Job sheet closed | What was done, any follow-up needed, when paperwork will arrive |
| Invoice sent | Invoice raised | Amount, how to pay, who to ask about it |
| Service or landlord check due | A set time before the due date | What is due, for which property, how to book |
| Quote sent | Quote issued | Where to find it, how long it is valid, how to accept or ask questions |
Keep each message to one purpose and one action. A confirmation that also asks for a review and advertises a boiler plan does three jobs badly. If a message needs a reply, say exactly what reply you want: 'Reply YES to confirm or call us to change the time.'
What to keep human
Some news should never arrive by template. The test is simple: if the customer is likely to be upset, worried or out of pocket, a person should tell them, by phone, and the written follow-up comes after the conversation.
- A delay of more than a few hours, or a cancelled visit, especially for someone without heating or hot water.
- A price change. Anything you said about the price that the customer relied on is a term of the contract, and a change to the information you gave before the contract is not effective unless the customer expressly agrees[4].
- An appliance classed as unsafe or turned off for safety reasons. The engineer explains on site, and the written record follows.
- A repair that has not worked, or a fault that needs a second visit at the customer's cost.
- Any complaint, and any reply to a complaint.
- Anything involving a vulnerable customer who has asked to be phoned rather than messaged.
Scripts for the human conversations
Short scripts help the office handle bad news calmly and consistently. Adapt these to your own voice:
- Delay: 'I'm calling because your engineer has been held up on an emergency. I don't want you waiting without knowing. The earliest we can be with you is two o'clock; would that work, or would you prefer tomorrow morning?'
- Price change: 'The engineer has found the problem is bigger than we expected when we spoke. Before we do anything, I want to explain the options and what each would cost, so you can decide.'
- Second visit: 'The part we fitted has not fixed it, and I'm sorry about that. We'll come back and we won't charge for the return visit to diagnose it properly.'
- Complaint: 'Thank you for telling us. Can you take me through what happened? I'll write down what you say and call you back by tomorrow with what we'll do.'
Writing automated messages people trust
- Use your business name as the sender, the same every time, so customers recognise genuine messages and spot fakes.
- Write like a person: 'Dan is on his way and should be with you around 10:40' rather than 'Your technician has been dispatched.'
- Give a time or a short window, not 'today'.
- Never include payment links in messages customers are not expecting, and never ask for card details by text.
- Tell customers how to reach a person in every message.
- Test every template on your own phone before switching it on, and read it as a worried customer would.
Service messages and marketing are different
A booking confirmation or reminder is a service message about a job the customer asked for. A text promoting a boiler service offer is direct marketing. Under PECR, direct-marketing emails and texts to individuals need prior consent, unless the soft opt-in for existing customers applies[2]. Keep the two separate: no offers tucked into reminders, and a clear way to opt out of anything promotional. Mixing them puts the trust you have built with service messages at risk.
Keep a person checking
Automation fails quietly. A template with the wrong arrival window, or a reminder that keeps firing for a cancelled job, can go out to dozens of customers before anyone notices. Most UK businesses that use AI already build in human checking: 84% reported at least some checking of AI outputs, and 67% significant checking[3]. Apply the same habit to any automated messaging. Someone should own the templates, review the message log weekly and handle every reply that is not a simple yes.
Customers' comfort with automated contact is mixed. Six in ten UK adults had used AI chatbots in the past three months, yet only around four in ten expected AI to have a positive impact on them, and three in ten a negative one[5]. Automated messages that are accurate and easy to reply to win people round; ones that trap them in a loop do the opposite. If you use an AI employee such as Ava on the phones, the same rule applies: it handles the routine call and the details, and a person takes over when the customer needs one.
A phased rollout
Month one: confirmations and reminders
Switch on booking confirmations and day-before reminders only. Read every reply for a fortnight to find gaps in the wording.
Month two: on my way and job complete
Add messages triggered by the engineer's job status. Train engineers to update status in the app as they go, or the messages will be wrong.
Month three: paperwork and due dates
Automate invoice notifications and service or landlord-check reminders, with the property address in each message.
Every quarter: review
Look at calls to the office asking 'when is the engineer coming?' and 'have you sent my certificate?'. If they have not fallen, find out which message is failing.
Example: what changed for one firm
Example: a fictional six-engineer firm in Southampton switches on confirmations, reminders and 'on my way' messages over two months, using the job software it already has. The office keeps a tally of calls asking about arrival times, which falls from several a day to a couple. One problem shows up quickly: reminders still go out for jobs cancelled by phone, because the office was not closing them in the system. The fix is a process change, not a software one. Price changes and delays stay with the office manager, by phone, with a short text summary afterwards.
Frequently asked questions
Which customer messages should a plumber automate first?
Booking confirmations and day-before reminders. They are factual, triggered by a clear event and cut the number of calls asking when the engineer is coming.
Can I put an offer in an appointment reminder?
Keep reminders as service messages. Promotional content turns them into direct marketing, which needs consent unless the soft opt-in for existing customers applies[2].
Should a price increase ever be sent by text?
No. Call the customer, explain the options and get their agreement before doing the extra work; a change to the price information you gave is not effective unless expressly agreed[4]. Confirm it in writing afterwards.
References
Office for National Statistics (2026). Business insights and impact on the UK economy: wave 165 dataset (AI tables) (opens in a new tab) ↑
legislation.gov.uk (The National Archives) (2003). The Privacy and Electronic Communications (EC Directive) Regulations 2003, regulation 22 (opens in a new tab) ↑
Department for Science, Innovation and Technology (GOV.UK) (2025). AI adoption survey report 2025 (opens in a new tab) ↑
legislation.gov.uk (The National Archives) (2015). Consumer Rights Act 2015, section 50 (opens in a new tab) ↑
Department for Science, Innovation and Technology (GOV.UK) (2024). Public attitudes to data and AI: Tracker survey (Wave 4) report (opens in a new tab) ↑