Almost every conversation I have about automation starts in the wrong place. Someone has seen a demo, or a competitor mentioned something at a trade show, and the question arrives as "should we be using AI for this?"
That question has no good answer, because it starts from the tool. The useful version starts from your week.
Here is the way I actually work through it, and it takes about ten minutes on a napkin.
Start with what leaves, not what annoys
The most annoying job in your business and the most expensive one are rarely the same thing.
Invoicing is annoying. Everybody hates it. But an invoice that goes out on Thursday instead of Tuesday costs you almost nothing, because the money still arrives. Meanwhile the call you missed on Tuesday morning cost you an entire job, and you never found out it happened.
So the first question is not "what do I hate doing". It is:
- Where does won work leave the business without me noticing?
- Where does paid-for attention leave without converting?
Those two questions almost always point at the same handful of things: calls that were not answered, enquiries that were not replied to, quotes that were never chased, and customers who were never asked for a review.
That is not a coincidence. Those are the four points where your business is losing something it already paid to get.
The order that works for most small businesses
If you want the short version, it is this, and it holds for most service businesses:
- Answer faster. Missed calls and unanswered enquiries.
- Follow up reliably. Quotes and estimates that go quiet.
- Ask for reviews. Every job, every time.
- Then internal admin. Data entry, paperwork, scheduling, reporting.
Notice that internal admin comes last, even though it is the thing most people ask about first. There is a reason for that.
Automating admin makes your existing work cheaper to deliver. Automating response and follow-up makes you more work. When you are small, more work beats cheaper work almost every time, because you have capacity and you have not yet filled it.
The exception is when admin is genuinely stopping you from taking on more. If you physically cannot accept another job because the paperwork would drown you, then admin is your bottleneck and it goes first. That does happen. It is just rarer than people think.
The arithmetic that decides it
Before you automate anything, do this sum. It is crude and it is enough.
- How many times a month does this thing happen?
- What does it cost you each time it goes wrong?
- What percentage of the time does it go wrong now?
A roofer who misses fifteen calls a month, converts one in four of the calls he does answer, and averages a few thousand dollars a job, is losing something enormous. The sum is not close, and the build pays for itself almost immediately.
A cabinet maker who gets six enquiries a month and answers all of them personally within the hour is losing nothing. There is nothing for automation to recover. Anyone who tells that person to buy a lead follow-up system is selling, not advising.
Most businesses are somewhere in between, and the sum tells you which side of the line you are on. That is the whole point of doing it.
The four honest reasons not to automate yet
I turn work down for these reasons more often than people expect, so they are worth stating plainly.
Your volume is too low. If the thing happens twice a month, a calendar reminder and a notebook do the same job for nothing. Do not spend four figures to replace a sticky note.
Your process changes every month. Automating a process you are still redesigning means paying twice. Let it settle first. If you cannot write down how it works today, it is not ready.
Being wrong is expensive and unavoidable. Anything where a mistake costs real money, promises a customer something you cannot deliver, or touches legal or medical territory, keep a person in the middle. Automate the typing around the decision, never the decision.
You are hoping it replaces a person. It will not. It takes away the repetition, not the judgement. If you buy it expecting to stop answering the phone entirely, you will be disappointed three months in, and I would rather say that now than take the money.
What "first" actually looks like in practice
The mistake I see most is trying to do everything at once. Someone decides to automate, and six weeks later there are four half-finished workflows, nobody trusts any of them, and the whole idea gets quietly abandoned.
Do one thing. Get it working properly. Live with it for a month.
One workflow that you actually trust is worth more than four you are suspicious of, because trust is what determines whether you leave it switched on. The second build is always easier than the first, and by then you will have a much better idea of what the second one should be.
A worked example
Take a plumbing business with three vans.
The owner's instinct is to automate invoicing, because his wife does it at night and hates it. Reasonable. But work through the questions:
- Invoices go out late maybe twice a month. Cost of lateness: close to zero.
- Calls ring out perhaps forty times a month while everyone is on jobs. Roughly a third of those never call back.
That is not a close call. The first build is a text back on missed calls, which costs less than the invoicing work would have, and recovers actual jobs rather than saving an hour of typing.
Invoicing still gets automated. It is just third or fourth, once the thing that wins work is running and paying for the rest.
How to work out yours
You can do this alone with the questions above, and plenty of people should. If you want a second opinion, the automation audit I offer is free, takes twenty to thirty minutes, and ends with one to three specific things worth building, real prices, and an honest note on anything I would leave alone.
It also ends with "not yet" reasonably often. That is not a sales technique. A business that automates too early gets a bad first experience and never tries again, and I would rather be the person you call next year than the one who took your money this year.
If you would rather read about the mechanics first, there is a plain-English breakdown of what AI automation actually is and what it costs, and a page on what I automate for contractors specifically.