The short version
In a trades business, two automations account for nearly all of the return and both are boring. The first is instant acknowledgement: every lead, from every channel, gets a real reply inside a minute and lands in one place a named person watches. The second is quote follow-up: four short touches over two weeks, because most shops send a quote and then send nothing at all. Everything else, the dashboards, the AI agents that route things, the scoring models, is a demo that adds maintenance and rarely adds revenue. Build these on the tools you already pay for, with credentials in your own name and a one-page diagram of how it works, so nothing switches off when a vendor relationship ends. And write down what happens today before you automate anything, because automating a broken intake just makes it fail faster and more politely.
The demo problem
Every automation pitch shows the same thing. A diagram with a dozen boxes, an AI node in the middle, and a claim about hours saved per week. It looks impressive and it almost never survives contact with a shop that has four vans and a dispatcher who also does invoicing.
The reason is that the diagram automates a process nobody has written down. In most contracting businesses the actual process is a set of habits distributed across three people and a whiteboard. Turning habits into a flowchart is the work. The software part is an afternoon.
So the first thing we do, before touching a tool, is spend two hours mapping what happens now. Every way a lead arrives. Who sees it. What they do. What happens when they are on a job. Where it gets written down. Owners often find the map more useful than the automation, because it is usually the first time anyone has seen the whole thing on one page, and it always contains at least one channel nobody is watching.
The first thing worth automating
Instant acknowledgement, and it is not close.
The gap between replying to a web enquiry in five minutes and replying in an hour is the difference between a conversation and a voicemail. People submit forms to three companies. The first real reply usually wins, not because they are the best but because the customer stops looking.
What we build is unglamorous: every form, every text, every chat enquiry triggers a reply within sixty seconds that has a person's name on it and says what happens next and when. At the same time, an alert goes to whoever is on duty, on their phone, with the address and the job type in the notification so they do not have to open anything to know whether it is urgent.
The part that makes it work is not the speed of the machine. It is the one place everything lands. Most shops have leads arriving in a website inbox, a Facebook page nobody has logged into since spring, a text thread on the owner's personal phone, and a voicemail box. One inbox, one owner, one rule for what happens when that person is on a roof.
The second thing worth automating
Quote follow-up, which nearly every shop does badly and many do not do at all.
The pattern is consistent. A quote goes out. The customer does not reply, because they are getting two more and life is busy. Nobody chases, partly because chasing feels like begging and partly because it is nobody's job. Three weeks later the work went to whoever followed up once.
Four touches over two weeks, each short, each asking something different. Day one, confirm it arrived and ask whether anything is unclear. Day four, answer the question everyone has and nobody asks, usually about timing or what would change the price. Day eight, one line offering to walk through it on the phone. Day fourteen, a plain ask: are you going ahead, and if not, it helps us to know why.
That last one produces the most useful information in the whole sequence. People will tell you they went with someone two thousand cheaper, or that they postponed until spring. The first answer is a pricing conversation. The second is a lead for March, and knowing which is which is worth having.
Write them in your voice. A follow-up sequence that sounds like software makes the quote feel like it came from software.
Things that look worth automating and are not
Lead scoring. In a business with one city and four job types, the dispatcher already knows which is urgent, and a score adds a number nobody trusts.
Dashboards. Almost every custom dashboard we have seen built for a shop this size is opened enthusiastically for two weeks and then never again. The numbers that matter fit in a weekly email.
Full customer journey nurture. Twelve emails over six months about maintenance tips, for a customer who will next need you in four years when the tank goes. The cost is not the build, it is that somebody has to keep it current, and nobody does.
An AI agent deciding when to contact customers. Rules and timers are predictable, explainable to a customer who asks why they got a text, and they do not surprise you at eleven at night. We use a model for narrow jobs like summarising a long voicemail or sorting an enquiry by type, and not for deciding who gets messaged.
Build it where your data already lives
The pitch you will hear most is a platform that does everything: CRM, pipeline, texting, email, funnels, automation, for a few hundred a month. The features are real. The problem is what happens at the end.
Your automation lives inside their system. Your pipeline history lives there. Stop paying and it all stops, and moving off it during a busy season is a project nobody has time for. That is leverage, and it is leverage against you by design.
The alternative is not harder. Most shops already pay for something that holds the data, whether that is Jobber, Housecall Pro or ServiceTitan, and the connective tissue can sit in a workflow tool on accounts in your own name. If you fire whoever built it, everything keeps running. That should be the baseline expectation and it usually is not.
What it should cost
A build like this is a week of work, and a week of work has a knowable price. We charge $1,250 once, because the scope is genuinely fixed: map, instant response, quote sequence, review request, one reactivation pass, and documentation.
The two alternatives are an hourly consultant, where the work is often good but the total is unknowable in advance, and a done-for-you platform retainer at three to five hundred a month forever for something that took a week to build.
There is a legitimate ongoing piece, and it is smaller than it is usually sold as. Somebody should check monthly that the sequences are still firing, that a software update has not silently broken a connection, and that the reply rates are not drifting. That is a couple of hundred a month of real work, it is optional, and nothing should switch off if you decline it.
The reactivation pass nobody expects
The item that surprises owners most is the cheapest. One clean pass over every quote from the last twelve months that went quiet.
Not a marketing blast. A short, plain message per person: we quoted you for this in March, it is still on file, is it something you are thinking about for this season. Some of those jobs never got done by anyone. A share of them come back.
We have watched this single step pay for an entire build in the first fortnight more than once. It works precisely because it is not automation in any impressive sense. It is a list somebody finally read.
Who owns it after we leave
Name one person in the shop as the owner of the automation before it goes live, the same way you would for a van. Not the most technical person, the person who notices when something stops. Their whole job is to say that the follow-up texts have not gone out since Thursday, which is the failure that otherwise runs silently for a month.
Give them the one-page diagram and the logins, and put a standing fifteen minutes in the calendar once a month to open the sequence and read the last ten messages that went out. That is the entire maintenance routine, and shops that do it keep getting value from the build three years later.
Check it on your own site
Map what happens now, on one page
Every channel a lead arrives through, who sees it, what they do, and what happens when they are on a job. Two hours. You will find at least one channel nobody is watching.
Time your current first response
Submit your own form on a Tuesday afternoon and again on a Saturday morning, then time the reply. Most owners are wrong about this number by an order of magnitude.
Put every lead in one inbox with one owner
Before automating anything. A faster process that still routes to four places just distributes the failure more efficiently.
Write four follow-up messages in your own words
Day one, four, eight and fourteen. Short, each asking something different, the last one asking outright whether it is going ahead and why not.
Keep every credential in your name
Accounts you own, access you can revoke, and a one-page diagram of how it fits together. If a vendor cannot hand that over, the automation is theirs and not yours.
Run the reactivation pass last
Once the new leads are handled properly, go back over a year of quiet quotes. It is the cheapest step and it frequently pays for everything above it.