Small businesses don't have an AI problem. They have an AI curation problem.

Almost every small business owner I talk to is using AI now. Not a little. A lot. ChatGPT for emails, Claude for captions, a separate tool for meeting notes, another for image generation, a fourth one their cousin recommended at Thanksgiving. Most of it on autopilot. Almost none of it questioned.

And underneath it, two feelings stacked on top of each other. One: I have to use this or I'll fall behind. Two: I know this stuff is bad for the planet and I'm trying not to think about it.

Both feelings are real. Neither one gets talked about honestly — especially together. So let's.

The actual problem isn't using too little. It's using too much, badly.

The story we keep getting sold is that small businesses are behind on AI and need to catch up. In my client conversations, that is not what I see. What I see is the opposite — owners running every email, every caption, every "can you make me a logo" through three different tools, accepting whatever comes out, and then wondering why their brand suddenly sounds like everyone else's.

Using AI more is not the goal. Using it well is. And "well" almost always means using it in fewer places, with more intention, and editing harder on the way out.

The honest part about the environmental cost

Every prompt has a real-world footprint. Water to cool the data centers. Power to run them. Hardware that took rare earths to build. It's not catastrophic per query, but it is not nothing — and the per-query cost is multiplied by the fact that we're all being nudged to use it for everything, all the time.

I'm not going to tell you to stop. I use these tools every day, and so do my clients. But pretending the cost doesn't exist while we 10x our usage is the kind of thinking that got us into a few other messes. The grown-up version: use it where it earns its keep, skip it where it doesn't, and stop treating "more AI" as a virtue.

Curation is the environmental move. Every prompt you don't send is a prompt that didn't run.

Start with the question, not the tool

The mistake almost everyone makes: they open ChatGPT, stare at the blinking cursor, and try to think of something to ask. That's backwards.

Start with the things in your week that are slow, repetitive, or that you put off because they're annoying. That short list is your AI to-do list. Everything not on the list doesn't need a tool — it needs you.

For most small businesses, the list looks pretty similar:

• Writing the same kind of email for the hundredth time
• Drafting a social caption when you'd rather be doing the actual work
• Summarizing a long PDF, contract, or onboarding doc
• Turning a messy voice memo into clean notes
• Cleaning up a spreadsheet someone sent you in a state

Notice what's not on that list. Strategy. Pricing. Hiring. Your About page. Anything where the wrong answer costs you real money or real trust. AI is a great intern. It is a bad CEO. It is also a bad copywriter for the things that have to actually sound like you.

The three places AI earns its keep

1. First drafts of things you'd rather not write

Client follow-ups. Newsletter sections. FAQ answers. A first draft from ChatGPT or Claude takes 30 seconds and gets you out of the staring-at-a-blank-page tax. Your job is editing, not writing — and editing is much faster.

One rule: never ship the first draft. AI-default copy has a specific smell — too tidy, too eager, lots of "in today's fast-paced world." Strip it out. Make it sound like you. If it doesn't sound like you, your customers can tell. (They already can, on a lot of small business feeds.)

2. Summarizing and explaining

Paste in a 14-page service agreement and ask for the five things that matter. Paste in a competitor's pricing page and ask what they're charging for what. This is where AI saves the most time per minute spent, and the stakes are low — you're reading a summary, not betting the business on it. Verify the parts that would actually cost you something to get wrong.

3. Turning a mess into a structure

You have ten pages of notes from client calls. You have a voice memo from a walk where you finally figured out your offer. AI is excellent at taking a pile and giving you a shape. You'll still edit it. But it's a much shorter trip from "pile" to "draft" than doing it cold.

Where it doesn't earn its keep — and where the cost isn't worth it

• A real logo for a real business. Generators give you a shape; you need an actual brand. (Different post.)
• Anything that has to sound exactly like you — your About page, a heartfelt email to a long-time client, a hard conversation with a vendor.
• Pricing decisions. AI will give you a confident range based on nothing.
• Legal, tax, or HR answers you're going to act on. Use it to understand the topic; don't use it to make the call.
• Customer service for anything that matters. A bad chatbot loses customers faster than no chatbot.
• "Just for fun" image generation, ten times a day, because it's there. That one is almost pure footprint, no payoff.

Two tools, not eleven

If you're already three subscriptions deep and not sure why: this is your sign to cut back. Pick one general assistant and, at most, one tool that fits a specific job. Cancel the rest for a month and see what you actually miss.

General assistant: I use Claude, and it's the one I recommend. In my experience it panders less, hedges less, and is accurate more often than the alternatives. It's also easier to teach — you can hand it your brand voice, your usual customers, the way you actually talk, and it will hold onto that instead of drifting back to generic marketing-speak after three prompts. (If you've never "taught" an AI before, that's something I walk clients through.) ChatGPT is fine if you already have it. The point is: pick one, pay for it, and go deep.

Specific tool: pick the one thing in your week that hurts the most — meeting notes, audio editing, image cleanup — and find a tool aimed at that. Don't subscribe to all of them. Pick one. Use it for a month. Then decide if it earned its spot.

Eleven tools is a hobby. Two tools is a workflow. Two tools is also a fraction of the footprint.

The anxiety part

The pressure to use more AI, faster, is being manufactured. A lot of it is coming from companies that need you to use more AI, faster, so their numbers work. You are allowed to opt out of the parts that don't serve your business — and you're allowed to do that without feeling like you're being left behind.

The owners who win the next few years are not the ones using the most AI. They are the ones who use it on purpose, in a few places, and put their actual attention into the parts of the business that need a human. The brand. The relationships. The judgment calls. The things customers can feel.

Use less. Use better. Edit harder. That's the whole strategy.