"Should I be using AI for this?" — a decision rule for small business marketing

The question I hear most from clients this year is some version of "should I be using AI for this". It comes up about blog posts, about email, about proposals, about replying to reviews. The honest answer is that it depends, and "it depends" is useless advice on its own, so I have settled on a rule I now give every client. It has held up well enough in real client work that it is worth writing down.

The rule: separate thinking from doing

AI is good at two very different jobs, and the most common mistake I see small businesses make is muddling them.

The first job is thinking. When I am unsure about a decision, I use ChatGPT the way I would use a colleague to think out loud with. What are my options for this campaign. What am I missing in this pricing decision. Argue against my plan. This kind of use is low risk and high value, because nothing goes anywhere near a customer. The output is better thinking, and you are still the one doing it.

The second job is doing. When I know what needs to be made, I use Claude to produce it. Draft the email sequence, build the report, turn the meeting notes into an action list, write the first version of the web page. This is where the real time savings live, and it needs more care, because the output does reach customers. Your facts, your judgement and your voice have to go over it before anything ships.

That is the whole rule. Reflect with one tool, produce with the other, and always know which mode you are in. Clients who keep the two jobs separate get value quickly. Clients who blur them end up either publishing half-thought ideas or spending an hour polishing a brainstorm that never needed polishing.

What I tell clients to skip

Where not to use AI matters just as much.

Skip it for the moments where the customer needs to hear from you personally. A response to a complaint. A thank-you to a long-standing client. An apology, a condolence, a difficult conversation about price. These moments carry the relationship, and readers can feel the difference even when they cannot name it.

Skip it for decisions only you can make. AI can lay out the options for your positioning or your pricing, and it does that well. It cannot know your appetite for risk, your capacity this quarter, or what you promised a customer over the phone last Tuesday. Use it to prepare the decision. The call stays with you.

And skip the tool-collecting. You do not need eleven subscriptions. One tool for thinking and one for doing, used consistently, will beat a drawer full of half-finished trials every time.

Why the human moments now matter more, not less

There is a bigger shift underneath this rule, and it is the thing I find myself saying in almost every strategy session. As AI absorbs more of the routine steps in your customer's journey, the confirmations, the follow-ups, the scheduling, the moments that stay human become the ones that decide whether people trust you, stay with you and refer you. Automation raises the stakes on the human touchpoints. It does not replace them.

That is why I pair every AI conversation with customer journey mapping. The map shows you which moments in the journey actually build trust and loyalty. AI and automation clear the routine work out of the way so you have the time and attention to deliver those moments properly. Automate the process, never the relationship.

A quick test before you hand anything to AI

When a client is unsure, I give them three questions. Is this thinking or doing, and am I using the right tool for that job. Does this moment need to be unmistakably me, or is it fine to shape a draft. And has a human, meaning you, checked every fact and read it aloud before it ships. If you can answer all three, use the tool with a clear conscience.

Where to from here

Rules like this one are most useful when they arrive at the moment you need them, which in a small business is roughly weekly. I share what I am seeing work with real clients, including the workflows behind it, in my newsletter. It is short, practical and written for owners who want AI to save them time without costing them their voice.