Most talk about AI marketing for small business is about the wrong question. People argue about whether AI is good enough to do marketing. That's not the useful question. The useful question is which parts of marketing are production and which parts are judgment.
Production is typing, formatting, resizing, summarizing, and making ten versions of something. AI is fast and pretty good at it. Judgment is deciding what to say, what to offer, who to trust, and what goes out with your name on it. AI is bad at that, and it doesn't know it's bad at it.
Here's my rule: let AI do the production, keep humans on the judgment. Everything below is that rule applied.
Production got cheap. Judgment didn't.
For years, the expensive part of marketing was production. A blog post took a writer most of a day. Ad variations took a designer an afternoon. Monthly reports took someone pulling numbers from five places. Agencies were built to sell that production, with layers of management on top.
AI changed the price of production. What hasn't changed is the cost of knowing what to produce. That still takes experience in your industry, your market, and your customers. It's why I argue that the agency model is dead: the layers were built to manage production, and production isn't the bottleneck anymore.
What should AI do in small business marketing?
Here's how I sort tasks for the service businesses I work with. Three buckets: AI does it, AI drafts and a human finishes, and keep it human.
| Task | Who does it | Why |
|---|---|---|
| Reformatting a post for Facebook, LinkedIn, and Google | AI does it | Pure production, low risk |
| Summarizing reports and call notes | AI does it | Fast, and you check the numbers anyway |
| Ad headline and description variations | AI drafts, human finishes | Great for volume, but claims need checking |
| Blog first drafts | AI drafts, human finishes | The structure is easy, the expertise isn't |
| Promo emails | AI drafts, human finishes | The wording is easy, the offer is yours |
| Replies to happy reviews | AI drafts, human finishes | Quick to approve, still public |
| Replies to angry customers | Keep human | Tone and facts matter too much |
| Strategy and priorities | Keep human | Needs context AI doesn't have |
| Offers and pricing | Keep human | Your margin, your call |
| Customer relationships | Keep human | People hire people |
| Approving anything public | Keep human | Your name is on it |
| Photos of your real work | Keep human | Never fake a job photo |
Bucket 1: Let AI just do it
These are the tasks where a mistake is cheap and obvious. Resizing and reformatting content. Turning a long report into five bullet points. Cleaning up a messy transcript of a call. Sorting a list of customer questions into themes.
Nobody needs a senior marketer for this. You barely need to check it. If you're still doing any of this by hand, start with these five jobs to hand to AI this week.
Bucket 2: AI drafts, a human finishes
This is where most marketing lives, and where most of the time savings come from. AI gets you from a blank page to 70 percent done. A person who knows the business takes it the rest of the way.
Ad variations are a good example. AI can write 15 headline options for a furnace tune-up ad in seconds. A human picks the four that sound like the business and catches the one that promises "same-day service" when you can't actually do that in January.
Blog drafts work the same way. AI can structure a post on "repair or replace your water heater." It can't tell the reader that in your experience the second repair call in a year is usually the moment to start pricing replacements. That sentence is the part that ranks and sells. I go deep on this in is AI content killing your SEO.
The test for bucket 2: if the finished piece could have been written about any company in any city, the human didn't finish it. Add the detail only you know.
Bucket 3: Keep it human
Strategy
Which services to push this quarter, which zip codes to target, whether to spend on ads or fix your website first. AI can list options. It can't weigh them against your crew's capacity, your cash flow, and the fact that your best tech is out for surgery in March.
Offers and pricing
AI will happily write "20 percent off all services!" It doesn't know your margins. Every offer should come from someone who knows what a job actually costs you.
Customer relationships
The follow-up call to a past customer, the conversation with a property manager who sends you 30 jobs a year. AI can remind you to do it and draft notes. The relationship itself is yours.
Angry reviews
A one-star review from an upset customer is a public conversation that future customers will read. An AI-drafted reply often sounds polished and hollow at exactly the moment you need to sound like a real person who cares. Write it yourself, or at least rewrite the draft heavily, and keep the private details private.
Approving anything public
Nothing goes out under your name without a human reading it. Not ads, not posts, not emails. This is the single rule that prevents almost every AI embarrassment I've seen.
Photos of real work
AI image tools can make a beautiful picture of a finished kitchen remodel. Don't use it as yours. Customers hire you based on what you've actually done, and a faked job photo is a lie that falls apart the first time someone compares it to the real thing. AI-made graphics for a tip or a holiday post are fine. Pictures that claim to show your work must be your work.
Where owners get this wrong
I see two failure modes, and they're opposites.
- Handing judgment to AI. The owner lets ChatGPT pick the offer, write the angry-review reply, and post straight to Facebook. It's fast, and it slowly makes the business sound like nobody's running it.
- Doing production by hand. The owner refuses to use AI at all and spends Sunday nights writing posts from scratch. The marketing is authentic, and there's almost none of it.
The fix for both is the same split. Push the production to AI and protect your time for the decisions.
A quick way to sort your own list
Write down every marketing task you touched last month. Then ask two questions about each one.
- If this is wrong, who notices? If the answer is "nobody, and it's easy to fix," AI can do it. If the answer is "a customer," a human finishes it.
- Does this need something only I know? Your margins, your crew's schedule, the customer's history, what the job actually looked like. If yes, it stays human, and AI can only help with the typing.
Most owners find that half their list is pure production they've been doing by hand. That's the half to hand off first. If you're not sure where to start, here's what to hand off first.
What this means for who you hire
If production is cheap, you shouldn't be paying production prices with management markup on top. What you want is one experienced person who uses AI for the production and spends their time on judgment for your business. That's the whole idea behind a fractional marketer instead of an agency or a full-time hire.
It's also how I work. I bill $125 per production hour, and because AI handles the typing, those hours go further. The judgment is what you're actually buying.
Want this handled for you?
$125 per production hour. Meetings are free. Month-to-month, and you own everything. Want a reference first? I'll connect you with someone I did similar work for.
Book a 30-minute call


