Stop Buying AI Tools. Get Someone to Run the Ones You Have.

Pull up last month's credit card statement and count the AI. The note taker. The writing assistant. The chatbot on the website. Two or three ChatGPT seats. The AI upgrade your CRM talked you into. When owners ask me what AI tools for small business cost, that's usually my answer: whatever is on that statement, plus the work none of them are doing.

Here's the uncomfortable part. In most businesses I look at, the problem isn't the tools. It's that nobody is assigned to run them.

How AI subscription sprawl happens

Nobody sets out to buy eight AI tools. It happens one reasonable decision at a time.

• The free trial that quietly converted. Someone tried it for a week, then forgot the card was on file.
• The upsell inside software you already pay for. Your scheduling tool, CRM or email platform adds an "AI" tier, and someone clicks upgrade.
• Everybody buys their own. The office manager has one assistant, the sales guy has another, and you have a third.
• The webinar. Somebody watched a demo that made one tool look like magic. It was magic for the person running the demo, who knew exactly what to ask it.

None of these are dumb purchases. They're just purchases without an owner.

What do AI tools for small business cost, really?

The fee is the easy part. For reference, Claude's Pro plan is >$20 a month billed monthly as of this writing. A per-seat fee like that isn't what hurts.

What hurts is everything around the fee:

True monthly cost of an AI tool = fee + unused seats + setup nobody finished + the output it was supposed to produce and didn't

That last piece is the big one. If you bought an AI tool to write follow-up emails and the follow-up emails still aren't going out, you're paying the fee and still eating the cost of the missing follow-up. The tool didn't fail. It just never got driven.

A worked example (hypothetical)

Say you run a 12-person property management company. Your statement shows:

That's about $313 a month. One line out of five is earning its keep, and two of its three seats are idle. The fix isn't a better tool. It's cutting the dead ones and putting someone in charge of the one that works.

Your numbers will be different. Run them anyway.

Why one capable agent beats ten point tools

A point tool does one job. An AI note taker takes notes. An AI writing tool writes. A year ago that made sense, because general AI assistants lived in a chat window and couldn't do much outside it.

That changed. General assistants can now operate the software you already have. >Claude in Chrome can read, click, type and fill out forms in your browser. OpenAI's ChatGPT Work has a >cloud browser that clicks through sites and pauses for your sign-in or approval. Both are on paid plans as of this writing.

So a lot of what you bought a point tool for is now just a task on a list. Draft the follow-up. Summarize the call. Update the listing. Build the report. One capable agent with access to your real tools can do most of it, if someone tells it what to do and checks the result.

That "if" is doing all the work in that sentence.

The missing piece is a driver

I've said this in the hammer post and I'll keep saying it: owning the tool doesn't make the work happen. Somebody has to:

• Decide what gets done this week, and in what order
• Give the AI the right access, without handing out passwords
• Write instructions clear enough that the output is usable
• Check every piece before it reaches a customer or goes live
• Notice what's working and what's just noise

That's a job. It's not your job, because your hour is worth more running the business. And it's usually not your office manager's job either, because they're already full.

The AI tool audit worksheet

Do this in one sitting. It takes about an hour.

1. List every AI charge. Credit cards, app stores, and the "AI" add-ons buried inside other subscriptions. Check with your team. Someone always has one you didn't know about.
2. Write the job each tool was bought to do. One sentence. "Write our follow-up emails." "Take notes on sales calls."
3. Find the last real output. Not the last login. The last time it produced something a customer saw or that changed a decision.
4. Name the driver. Who is responsible for running it every week? If the answer is "nobody," write that down.
5. Ask: could your main assistant do this job? If a general assistant with access to your existing software could do it, the point tool is a candidate to cut.
6. Sort into three piles. Keep (it has a driver and real output). Merge (its job moves to your main assistant). Cancel (no driver, no output, no plan).

If your non-AI software has the same problem, the subscription graveyard post runs the same math on the rest of your stack.

Prompts to run the audit faster

Export your card transactions as a CSV first. Delete the account number column before you upload anything. Never paste a full card number into an AI chat.

Prompt: Find the AI subscriptions in my statement

Attached is a CSV of my business card transactions for the last [3] months. I removed the account numbers.

Find every recurring charge that is an AI tool or an AI add-on to another tool. For each, give:
- Vendor name as it appears on the statement
- Monthly amount (average if it changed)
- Your best guess at what the tool does, marked "guess" if you're not sure

Do not include one-time purchases. Total the monthly AI spend at the bottom. If a charge is ambiguous, list it separately as a question for me.

Prompt: Can my main assistant cover this tool's job?

I pay for [tool name] to [job it was bought to do]. My team also has [Claude / ChatGPT] on a paid plan, and we use [list your main software, e.g. QuickBooks, Google Workspace, our CRM].

Tell me:
1. Which parts of that job a general AI assistant could do with access to our existing software.
2. Which parts genuinely need the specialized tool, and why.
3. What we would lose by cancelling it.

Be blunt. If the specialized tool is clearly worth keeping, say so. Do not recommend any new tools.

What to keep

After the audit, most businesses I see land in about the same place:

• One general assistant for each person who uses it every day.
• AI features inside your core software only if someone uses them weekly.
• Nothing new until there's a named driver and a clear job.

That's a smaller bill. More important, it's a bill attached to output.

Frequently asked questions

How much should a small business spend on AI tools?

There's no right number, but the order matters more than the amount. Start with one general AI assistant for the people who will use it daily. Add anything else only when someone is assigned to run it and you can name the output it produces each week. A tool nobody drives costs its full price and returns nothing.

Which AI tool should a small business buy first?

A general assistant like Claude or ChatGPT on a paid plan, for the person who will actually use it. Those can now draft, analyze, and on some plans work inside your browser. Most single-purpose AI tools do a narrow slice of what a general assistant can do once someone sets it up properly.

Are the AI add-ons inside my existing software worth it?

Sometimes. If your CRM or accounting tool offers an AI feature that runs where your data already lives and someone on the team actually uses it weekly, it can be worth the upgrade. If nobody has opened it in a month, it goes on the cancel list like anything else.

What does it cost to have someone run the AI tools for me?

With me, it's a flat $499 a month, month to month, with no setup fee. That covers my AI agents working your task list, my review of every piece of work and a weekly meeting. The first month is $99. You keep ownership of every account and tool.

Cancel one tool this week

Run the worksheet and cancel the first tool with no driver and no output. Then, if you want someone to actually run what's left, that's what I do. The details are on my pricing page, and the first month is $99 here.