Automate the inbound phone first, in most cases, because a missed call is a job lost to silence rather than to a competitor's better pitch, and the loss is happening today whether or not you measure it. Everything else on the usual list, quoting, scheduling, invoicing, paperwork, is a slower leak that you can see.
That is the default answer. Below is how to check whether it is the right answer for your shop, because sometimes it is not.
What is the actual test for choosing first?
Rank candidates on three things: how often the task happens, how much money leaks each time, and how easily you could tell whether the fix worked. The first automation should score high on all three, not just on annoyance.
Annoyance is the wrong criterion and it is the one most owners use. The most irritating task in a trades business is usually paperwork, but paperwork loses hours, and the phone loses jobs.
| Candidate | Frequency | Money per miss | Measurable in 30 days |
|---|---|---|---|
| Missed inbound calls | Daily | A whole job | Yes |
| Quote follow-up | Weekly | A whole job, delayed | Yes |
| Invoice chase | Monthly | Cash flow, not margin | Yes |
| Scheduling confirmations | Daily | A wasted truck roll | Yes |
| Field paperwork | Daily | Unpaid admin hours | Harder |
| Job costing | Per job | Invisible margin erosion | No, takes months |
Why not automate job costing first, when that is where margin disappears?
Because you will not know for months whether it worked, and a first automation that cannot be judged quickly will not survive contact with a busy season. Job costing is often the biggest prize and it is almost always the wrong opening move.
Start with something that either works or does not within thirty days. Win first, then go after the harder number with credibility and a team that believes the last one worked.
When is the phone not the right first automation?
When you already answer everything. If you have a full-time office person who reliably picks up, and your call abandonment is genuinely near zero, the phone is not your leak and automating it buys nothing.
In that case the next candidate is normally quote follow-up, because a quote that goes quiet is a job you already paid to win. Sales effort is sunk by the time the quote goes out.
How do I find out where my money is actually leaking?
Count three numbers over two weeks: calls that went unanswered, quotes sent with no reply and no follow-up, and days between job completion and invoice sent. It costs nothing but attention and it usually settles the argument.
If you cannot count them because the data lives in four places, that is itself the finding, and it is the most common one we deliver. A diagnostic exists for exactly that case.
Should I automate anything before fixing the process?
No, and this is the one rule worth being rigid about. Automating a process nobody agrees on just makes the disagreement happen faster and blames the software.
Data and systems have to be solid before AI does anything useful. Build on a shaky foundation and everything you build with AI becomes a house of cards. Fixing the process usually costs nothing except a decision somebody has been avoiding.
What is a realistic first project?
One workflow, live inside a month, with one number attached to it. Missed call text back is the most common because it changes nothing for the crew, keeps your published number, and either recovers jobs or does not.
Resist bundling. A first project with four workflows in it has four ways to fail and no clean read on which part worked.
What if the honest answer is that I should automate nothing yet?
Then that is the answer and you should hear it before you spend, not after. Most businesses are not ready for AI yet. That is not a criticism. It is the reality.
A shop with no documented process, no agreed source of truth and nobody to own a new system will not get value from automation regardless of what it buys. Fix that first, and it is usually free.
Part of our guide to business automation for Canadian trades.