What AI already does in an installation business: concrete tasks, no promises
What does AI already do in an installation business? Five concrete tasks, from draft replies to form checks, and an honest look at what AI does not do.
In this article
Short answer
AI already helps an installation business with five concrete tasks: a draft reply to an email, the urgency of conversations, checking submitted work orders, questions about your own figures and a daily brief. The AI proposes and prepares. A colleague decides, sends and stays responsible.
By the makers of stiptUpdated October 28, 20266 min read
What does AI really do in an installation business?
Many AI promises are about what might be possible one day. This is about what already works in daily work: answering email, deciding what comes first, checking work orders. It is preparatory work that costs colleagues hours. What AI does not replace is judgement.
The examples below come from stipt, where the AI colleague is called dot. The inbox helper, the urgency of conversations and the form check are off by default: an admin switches them on in Studio. If you use another package, check whether it does the same, and above all whether a person approves the result before anything goes out.
Five tasks AI already helps with
dot finds the record, reads what sits behind it and prepares a draft, with a short summary and suggestions such as a task or a planning link. You open the draft in the reply box, adjust it and send it.
Every morning dot reviews the open conversations and sets a red, yellow or grey flag, with the reason attached. The longer a customer waits, the higher the urgency climbs. dot leaves a colleague's flag alone.
A submitted form is checked right away, photos included: is the photo usable, does the serial number match the rating plate. The reviewer gets a verdict and the points to look at.
Ask in plain language how many quotes are open or how many appointments did not go ahead. dot calculates read-only and adds a chart where it fits. This is available to admins and colleagues with analytics access.
A personal overview at the start of the day of what needs attention: conversations that need an answer today, quotes about to expire, records that have stalled and your open tasks.
How an email goes from arrival to reply, with AI as the preparer.
- 1The email arrives
A customer writes about a breakdown, an appointment or a quote.
- 2dot finds the record
Through the sender, a stipt ID or other number in the email, an email address or phone number in the text, or postcode and house number. If the customer is certain, dot links the conversation itself. If he is only found by name, it is a suggestion, because it could be someone with the same name.
- 3dot reads what sits behind it
Status, appointments, forms, quotes, invoices and open tasks on that record.
- 4dot prepares it
A summary, a draft and one-click suggestions: create a task, change the status, send a planning link or mark it handled.
- 5A colleague decides
You read the draft, adjust it and send it yourself. A suggestion you carry out with a click, or you skip it.
- 6dot learns from it
You do not have to tell dot anything: what you carry out, dismiss or rewrite is the lesson. If something is not right, say so with the thumbs down. A lesson with text from an email or draft only counts once an admin switches it on.
Who does what?
| Task | The AI | A colleague |
|---|---|---|
| Answering email | Finds the record, summarises and writes a draft | Reads, adjusts and sends |
| Setting urgency | Reviews open conversations and sets a flag with the reason | Takes red first and can change or remove any flag |
| Checking a work order | Reviews form and photos and gives a verdict | Accepts or rejects |
| Looking up figures | Counts and calculates read-only, with a chart | Draws the conclusion and decides what happens next |
| Starting the day | Puts what needs attention in order, with the reason | Chooses what to pick up |
How far dot goes is set per tool. For the inbox helper and the form check the default is preparing: a colleague decides.
What does AI not do?
Honest about the limits, because that is where the disappointment starts.
AI does not send anything to a customer by email on its own. A draft stays a draft until a colleague sends it, and a suggestion to change a status or create a task you carry out with a click by default. How far dot goes you choose per tool. For the form check the default is that dot only prepares an analysis for the reviewer. If you want him to accept what is in order himself, you choose that deliberately. For the inbox helper you choose between watching along, preparing tasks and handling conversations itself.
AI also does not schedule anything through the chat, that stays with your planner. AI does not promise prices or discounts: that is a default limit you keep or adjust. And In its drafts AI works from what is in the record, the conversation and your knowledge base. The WhatsApp bot for customers therefore gives no answer without knowledge and hands over to a person when in doubt. That bot is off by default. It can only reschedule or cancel an appointment if you switch that action on, and only after the customer taps Confirm in WhatsApp.
How to start small
- Start with one task, for example draft replies, and leave the rest for a while
- Fill the knowledge base with what customers often ask: your website, your terms and your frequently asked questions
- Keep training mode on and give feedback for a few weeks on what the AI suggests
- For the form check, start with the mode where only the reviewer sees the verdict
- Put senders that never need attention, such as suppliers and parcel services, on the skip list
- After a month, look in the log at what the AI did and how often your colleagues agreed with it
Frequently asked questions
Can AI email customers on its own?
Not for email: the AI prepares a draft and a colleague sends it. On WhatsApp a separate customer bot can answer on its own, but it is off by default, draws only on your own knowledge and the record details you release, and hands over to a person when in doubt.
What if the AI gets it wrong?
You notice when you proofread, because a colleague sees every draft and every suggestion before it goes out. You give feedback with a thumb and a reason, and the AI uses those corrections as a lesson for next time; a colleague's lesson counts once an admin switches it on. The lessons can be found and switched off one by one.
What does the AI need to work?
A connected mailbox or WhatsApp, well-kept records and a knowledge base with what you know yourself: your website, terms and frequently asked questions. Without a knowledge base dot bases a draft only on the record and the conversation, and the WhatsApp customer bot gives no answer. The better your records, the better the draft.
What happens to my customer data when AI reads along?
Ask your supplier where the AI model runs, which parties process messages and whether they are used to train models. That belongs in the data processing agreement. The article on the GDPR and customer data in software goes through those questions.
Does AI replace my office staff?
The examples above take over preparatory work: looking things up, summarising, writing a first version and checking. Judgement, customer contact and agreements stay with people. Your colleagues end up with more time for the work that does need someone who knows the customer.
Read more
More articles. On software and processes in the installation trade.
See what stipt can handle for your business?
We'll show you what your working day looks like in stipt, set up for your kind of work and your way of working.