The First Thing to Automate, and How to Know It Worked

Adopters Pick Few Things, So Picking Right Matters More Than Picking Many

There is a quiet finding in the Census Bureau's work on how AI actually spreads through American businesses. Among firms that use AI at all, 57% use it in three or fewer business functions. Not thirty. Three or fewer. The picture in the trade press, where a shop rolls out an intelligent layer across every corner of the operation, does not describe what firms do. They pick a couple of things and stop.

That changes the question entirely. If you were going to automate twenty processes, order would barely matter, because you would reach all of them eventually. Since you are realistically going to automate two or three and then get busy again, the first one carries most of the weight. Get it wrong and you have spent your one credible attempt on something that did not move a dollar, which is how owners end up saying automation did not work for them.

The same research shows the most common function firms land on is sales and marketing, at 52%. That instinct is correct. The front of the operation is where the undecided money sits.

The Triage Test: Rank Your Leaks by Dollars at Risk

Every field service business has four candidate leaks, and owners tend to rank them by irritation rather than by dollars. The missed inbound call. The quote that does not go out the same day. The scheduling scramble. The invoice that ages sixty days.

Rank them by what is still undecided. A scheduling mistake is annoying, but the customer already chose you. A late invoice is a cash-timing problem on revenue you have already earned. Both are real, both cost you, and neither one is a decision the customer is still making.

The missed call and the slow quote are different in kind. That is a prospect actively choosing between you and whoever picks up next. A separate study in the same HBR article, tracking 1.25 million leads, found firms contacting a lead within an hour were nearly 7x more likely to qualify it than firms that tried an hour later, and more than 60x more likely than those who waited a day. That study is from 2011, and I keep citing it because nothing since has made owners faster. If anything, the customer's patience got shorter while the average response time stayed where it was.

Scheduling and invoicing move money you already won. Response speed moves money you are currently losing.

The Smallest Shops Are the Least Automated and the Least Ambitious About It

Here is the uncomfortable part. Firms under twenty employees sit at about 18% AI use, while firms with 250 or more employees jumped to 31%. The gradient is not surprising. What is surprising is the forward expectation: the very smallest shops, one to four employees, expect to reach only about 21% six months out. They are not planning to close the gap. They are planning to stay roughly where they are.

Construction sits among the lowest-adoption sectors in that same data, and 91% of specialty trade contractor firms have fewer than twenty employees. So the modal American trade shop is small, in a low-adoption sector, and not expecting that to change.

I do not read that as a technology problem. The tools are not hard. Nobody told these owners what order to do things in, so the whole category reads as an expensive, vague, all-or-nothing project that has to wait until the season slows down. It never slows down.

Get Your Own Baseline Before You Change Anything

Most 2-to-15 crew shops we have watched pull their own phone logs land somewhere between 20% and 40% of inbound calls unanswered, and almost none of them believed it before they pulled the report. Do not take that range as your number. Pull thirty days off your carrier's call log and compute it yourself. That figure is your baseline, not ours.

Do the same for quotes. Count how many went out the same day the request came in, over the last thirty days. Nobody publishes a trustworthy benchmark for that, and you do not need one, because the only valid comparison is against your own prior month.

Write both numbers down somewhere you will find them again. Half the value of this exercise is that in five weeks you will have something honest to compare against instead of a vague sense that things feel better.

What the Fix Actually Looks Like

The mechanism is not complicated and it is not the point of this post. On Plus, you get a dedicated phone line so inbound calls route somewhere accountable instead of a cell phone in a truck cab. On Pro, the AI Voice Agent answers when nobody can, captures the job details, and drops them into the CRM as a real record rather than a voicemail you will return at nine that night. Estimates go out from the same system, and the follow-up sequence runs whether or not you remember to run it.

If you want the longer version of why the unanswered call is such an expensive habit, The Missed-Call Economy walks through it. For how to judge any of this without taking a vendor's word, including ours, AI Tools for the Trades: Proof Over Hype is the more skeptical read.

How to Know It Worked

Two numbers, thirty days, compared only to your own baseline.

Median time to first response on inbound leads. Not average, because one 3 a.m. web form answered at 7 a.m. drags an average around. Median. If yours was four hours and it is now under ten minutes, that is the result. If it is still four hours, whatever you turned on is not in the path of real inbound volume, and you should find out why before you buy anything else.

Share of inbound calls answered, by a human or an agent, rather than going to voicemail. If your baseline was 70% and you are now at 92%, you can count the difference in jobs.

Then the cost side. Core is $79 a month on annual billing with two seats, Plus is $159 with four, Pro is $239 with six, and extra seats are a flat $25. Put your own average ticket next to that number. For most shops, one recovered job a month covers the subscription several times over, and you will know within thirty days whether you are getting it. If you want to see how the pieces fit before committing, the platform overview lays it out.

The Honest Take

This post is not for the single-operator shop, and that shop is the majority case. With 322,000 specialty trade firms under five employees, most owners reading this answer their own phone from the truck. If you take eight calls a week and miss one, your problem is not response infrastructure. It is that you are one person. Buy back time some other way first.

It is also not for a shop with a demand problem. Labor demand in the trades is growing, with HVAC mechanics projected to grow 11% through 2035 and plumbers 7%, but that is national, not your zip code. If the phone genuinely does not ring, faster answering of nothing is still nothing. Your first move is lead generation, and that is a different post.

The shop this is for runs two to fifteen crew, gets more inbound than one person can catch during a busy week, and has never actually measured what falls through. If that is you, pull the call log this week. The number will be worse than you think, and that is good news, because it means the fix is available.