“AI automation” is sold as a single thing you buy. It is not. For a trades or service business it is four or five specific jobs that happen to use similar technology underneath, and they have different costs, different payoffs, and no dependency on each other.
Treating them as one purchase — increasingly sold under the name AI employee — is how businesses end up paying for a platform to solve one problem. Treating them as separate jobs lets you buy the one that is costing you money and ignore the rest, possibly forever.
For transparency: Reign AI builds several of the things described here, so take the recommendations as coming from somebody with an interest. The order below is arranged by what usually pays back first rather than by what is most profitable to sell, and the last section is about when the answer is to buy nothing.
The jobs, honestly
1. Answering the phone. The most mature of the four and usually the first worth doing, because the loss is immediate and measurable: an unanswered call from someone with an urgent problem is frequently gone inside the hour. Whether it is worth solving for you is a calculation rather than a feeling, and what missed calls actually cost sets out the method and where each input comes from. The mechanics of what actually happens on such a call are in how AI receptionists work.
2. Following up after the call. Much less discussed and, for businesses that quote work, frequently worth more. Most service businesses make one follow-up attempt on a quote and then stop — not from laziness but because the person who would do it is on a roof. Everything about what a sane follow-up sequence looks like, and where it crosses into being a nuisance, is in lead follow-up after a missed call.
3. Running paid lead channels. Google’s Local Services Ads and ordinary Google Ads are two different products with two different money models, and most trades businesses are pitched them interchangeably. The distinction, and the part about who you actually pay, is in Google Ads versus Local Services Ads. Running them well is three further jobs: keeping Local Services settings honest now that lead credits are automated, pruning the searches a Google ad pays for with negative keywords, and tracking which calls became work.
4. Getting the work into a calendar. The least glamorous and the one that quietly determines whether any of the others produce anything. A booking that requires three phone calls to confirm is not a booking. What has to be true of your calendar before automated booking can work at all is in booking a job straight into your calendar.
There is a fifth that gets sold hard and is worth naming so you can recognise it: generating marketing content. It is real, and it is nowhere near the top of this list for a service business whose problem is that the phone rings while they are under a sink.
The order to do them in
Not the order above — the order your own numbers give you. Two measurements decide it, and both are free:
- Pull ninety days of call records. Answered versus unanswered, with timestamps. If the misses are large and clustered, job one is your job one. If you answer almost everything, skip it entirely — this is the single most common case where a business buys something it did not need.
- Count your dead quotes. Go back three months and count how many enquiries received exactly one follow-up attempt and then nothing. If that number is meaningful and your average job is worth having, job two is your job one.
Whichever number is bigger, start there. Do one thing, run it for a month, and look at whether anything changed before touching the second. That sounds slow and it is the only way to find out which change did what.
What does not work
Worth saying plainly, because these are the failures that recur:
- Automating a process that does not exist. If you have no follow-up habit at all, an automated sequence will send messages nobody planned and you will discover the plan by reading customer replies. Write down what should happen first, on paper, then automate that.
- Buying a bundle for one problem. A platform priced for four jobs is poor value when one of them was losing you work — and bundles are also where per-client setup work quietly becomes your job rather than the vendor’s.
- Automating the thing customers value most about you. In a referral-driven business the first conversation is often where the relationship starts. Overflow coverage is a different proposition from handing over every call, and the second one can cost more than it saves.
- Expecting any of it to create demand. All four jobs handle demand that already exists. None of them generate it. If the phone is quiet, this entire page is the wrong page.
When the answer is to do nothing
There is a real and unfashionable case for changing nothing, and it applies more often than the marketing in this category suggests:
- You answer most of your calls, and the ones you miss ring back.
- Your work comes from repeat customers and referrals rather than from people shopping.
- You are at capacity. Capturing more enquiries when you cannot service the ones you have is not an improvement, it is a longer list of people you will disappoint.
That last one is worth dwelling on, because it is the case most often argued past. If you are turning work away, better lead capture does not help you — pricing does.
Measure first. The measurement is free, it takes an afternoon, and it is the only part of this page that is true about your business rather than about businesses in general.
Common questions
Do I need to automate everything at once?
No, and trying to is the most common way this goes badly. These are independent jobs, not a platform, and each one can be adopted on its own and judged on its own. Doing them all at once means you cannot tell which one worked, you are changing several customer-facing things in the same week, and if something goes wrong you have no idea which change caused it. Pick the one that is costing you most right now and leave the rest alone.
What is the single thing most service businesses should do first?
Measure, before automating anything. Pull ninety days of call records and look at when the misses cluster, then look at how many enquiries got exactly one follow-up attempt and then nothing. Those two numbers decide the order of everything else, they cost an afternoon to get, and for a real share of businesses they show the problem is smaller than assumed — which is a legitimate result and saves a subscription.
Where does AI genuinely not help a trades business?
Anywhere the value comes from judgement or from being physically present. Pricing a job that has not been seen, deciding whether a repair is worth doing, managing a crew, handling a customer who is genuinely upset, and every part of the actual trade. It also does not create demand — if the phone is not ringing, better handling of the calls you get will not fix that, and the money belongs in being findable instead.
Is this different from the software my competitors advertise?
The underlying jobs are the same ones every vendor in this space sells, which is worth knowing when you are being pitched a platform. The useful question is not which product has the most features, it is which of these four jobs you actually have a problem with — because a bundle priced for all four is poor value when only one of them was losing you work.