You open ChatGPT and type something like "write me ten Facebook ad headlines for a plumbing company." Ten headlines come back in about four seconds. They are grammatical, they are on topic, and two of them are usable.
Here is the part nobody mentions. A plumber three states over typed nearly the same sentence that morning and got nearly the same ten. So did the intern at the agency pitching your competitor. The tool has no idea which of you it is talking to, so it does what it does when it knows nothing in particular: it returns the middle of everything it has ever read about plumbing ads.
That is the one job where AI output collapses toward average, and it is the job almost every owner points it at first.
Meanwhile there are four or five things you are genuinely stuck on, that you have been stuck on for years, where the same tool is quietly excellent. Almost nobody is pointing it there.
Everyone Is Selling You the Making
Look at what is actually being served to business owners right now. The most-watched videos are all the same shape. Nine ways to use AI in your business. Ten ways. Here are the tools, here is where to click, here is the free trial.
One of the better ones opens with the smartest idea in the entire genre. AI, it says, is not a strategy at all. It is a multiplier. If you are not already clear about what you sell and who you sell it to, it will not supply that clarity, it will only get you somewhere wrong faster.
Then the video spends the next twenty minutes on ten ways to move faster.
I am not being unfair to the people making these. They are mostly honest, and some of the tactics are fine. But notice what they have in common: every one of them ends at something you can buy. There is a subscription attached to making things and there is no subscription attached to thinking more clearly. So the making gets all the airtime, and the owner comes away believing AI is a machine that produces marketing assets, when the more valuable machine was sitting right next to it the whole time.
You Are Sitting on Data You Have Never Read
Log into your Google Ads account and open the search terms report for the last ninety days. Not the keywords you chose. The actual phrases people typed before they clicked.
You have probably never read it. Neither has most of the industry, and the reason is not laziness. It is that the thing runs to thousands of rows, most of them carrying one or two clicks, and reading it properly is a genuinely tedious afternoon that produces nothing for the first two hours.
That is the exact shape of work this technology is good at. Not writing. Reading.
Export it and ask a real question. Here is ninety days of search terms with cost, clicks and conversions against each one. What patterns are in here. What am I paying for repeatedly that has never once produced a lead. What are people looking for that I do not sell. What should I be asking my agency about.
The same applies to anything else you have been avoiding. Your last two years of monthly numbers. The CRM export nobody has opened since it was set up. Three spreadsheets that do not talk to each other.
I watched a demonstration of this recently that stuck with me, and I want to be straight that it was a product advertisement rather than a neutral test. Somebody uploaded three unrelated sales spreadsheets and asked why revenue dropped last month. The answer that came back was not "sales were down" or "your reps underperformed." It was that two large deals had been pushed into the following month, that both belonged to the same segment, and that the previous month had been unusually strong, so the drop was mostly timing.
That is a real answer. It is also the kind of answer that takes a person most of a day to find, which is precisely why nobody in a small business ever finds it. The tool in that video does not matter. The habit does.
Your Customers Already Wrote Your Ad Copy
Somewhere you have a pile of reviews. Support emails. Notes from sales calls. The questions people ask right before they buy, and the ones they ask right before they do not.
Feed the lot of it in and ask what words these people use to describe the problem before they found you. What are they afraid of getting wrong. What did they try first, and why did it not work.
This looks like asking AI to write your copy. It is the opposite. It is asking it to read two hundred reviews properly, which no human being in your business is ever going to do, and hand back the six things that keep coming up in the customer's own language. You are not inventing a message. You are retrieving one you already own and have never been able to see, because you were too close to it and there was too much of it.
Good advertising has always worked this way. The lines that land are the ones lifted off a real person, not the ones composed at a desk.
Make It Argue Against You Before You Spend Anything
You are about to commit to something. A new channel, a rebuild of the website, six months of budget at a number that makes you slightly uncomfortable.
Give it the whole picture. Here is my offer. Here is my margin, my close rate, what a customer is worth over two years, and the monthly number I am about to commit for six months. Now make the strongest possible case that this fails.
Then push harder. What would have to be true for this to work. Which of my assumptions is doing the most load-bearing. What is the cheapest thing I could do first to find out whether I am wrong.
Owners almost never get this. Your spouse wants to support you. Your staff have an interest in your optimism. The agency pitching you is, at that moment, selling. Nobody in the room is being paid to say no, and the one voice that will happily argue the other side costs nothing and is available at eleven at night, which is usually when you are actually thinking about it.
The Arithmetic You Have Been Putting Off
There is one number that decides most of your marketing, and most owners do not have it written down: what you can afford to pay to get a customer.
It comes out of your own figures. What you make on an average job, what proportion of leads become customers, what a customer is worth if they come back. We have written about how to derive it properly in Your First Ad Budget Isn't an Investment. It's Tuition., and the method has not changed. What has changed is that the excuse for not doing it is gone.
Give it your real numbers and ask it to work the ceiling out, show every step, and tell you which assumption the answer is most sensitive to. Then argue with it. You will find yourself correcting the close rate you gave it, which is the moment you learn you did not actually know your close rate.
You Cannot Audit the Work. You Can Ask Better Questions.
If somebody else runs your advertising, you already know you cannot judge the work directly, and you should not try. We made that case in How to Tell If the People Running Your Ads Are Any Good. The quality of the explanation you get is the real test, not the account itself.
This is where AI closes a gap that has been open for as long as our industry has existed. It has never seen your account and it cannot tell you whether your account is well run. But it knows the shape of the work extremely well, and it will hand you the questions.
Tell it what you pay, what you sell, who you sell to and what platform you are on, and ask what a competent operator should be able to answer without reaching. You will get a better list than most owners walk into a review meeting with. Then listen to how the answers come back. Somebody doing the work properly will be pleased you asked. That reaction is worth more than any of the individual answers.
None of this makes you a media buyer, and it is not meant to. It makes you a client who is hard to coast on, which is a different and more useful thing to be. It is also, frankly, the kind of client we would rather have. If you want to see how we answer those questions, that is what our paid search work looks like.
Where It Will Lie to You
Four failure modes, all of which I have watched cost somebody money.
It invents numbers and delivers them in exactly the same confident tone as the real ones. Industry benchmarks are the worst offender, because a plausible-sounding average cost per lead for your trade is the easiest thing in the world to generate and the hardest to check.
It agrees with you. Ask whether your plan is a good one and you will usually be told that it is, with three supporting reasons. You have to explicitly ask for the case against, which is why that section above is phrased the way it is.
It will be confident about an account it cannot see. If you have not given it the data, it is describing advertising in general, not yours.
And it does not know what happened in your market last month, or that your best competitor closed, or that your one good installer quit, unless you tell it.
The working rule is simple. It is reliable when it reads what you give it and reasons about that. Anything it produces that you did not supply, treat as a claim to be checked, not a fact.
The Part That Is Still Yours
There is a category of decision no amount of context will hand over.
Which customers you actually want, as opposed to the ones who are easiest to get. What you will not sell even though it would be profitable. Whether you want a business twice this size or the same size with less noise in it. What a bad quarter means at your kitchen table.
Those are not gaps in the technology that will close with the next release. They are the questions that make it your business rather than a business, and they were always going to stay with you.
Everything upstream of them, though, the reading and the arithmetic and the arguing and the preparing, is work you have been doing badly or not at all because there were not enough hours in the week. That part is now cheap.
Where to Start
Pick the file you have been avoiding. Ninety days of search terms, or last year's numbers, or the reviews. Upload it and ask what is in there that you would not have noticed.
You will not get a marketing strategy out of it. You will get one or two things you did not know about your own business, which is considerably more than you got from the ten headlines.
