The median small business now runs five AI tools and spends about $2,200 a year on them. Only 14 percent have wired any of it into how the work actually gets done. Those two numbers come from the same 2026 survey data, and read together they describe a problem most owners feel but few have named. You did not fall behind on AI. You bought too much of it, too fast, and never checked whether any of it pays.
This is the opposite of the panic story the vendors sell. The Census Bureau’s Business Trends and Outlook Survey, refreshed again on June 4, still puts AI use under 20 percent for firms with fewer than 20 employees. So you are early. But early and sloppy is its own trap. A subscription you forgot about bills you whether you log in or not. This week, before you start another free trial, run a one-hour audit of what you already pay for and cut what does not earn its line item.
List every tool, then put a dollar figure on each one
Open a spreadsheet. One row per AI tool, paid or free, that touches your business. Pull the paid ones straight from your bank and card statements for the last three months, because the ones you forgot about are exactly the ones quietly costing you. Add the free ones your team uses too, since a tool nobody owns is a risk even when it costs nothing.
Give each row five columns: tool name, what job it does, monthly cost, who on your team owns it, and the date someone last used it. That last column is the one that stings. Most owners running this audit find at least one tool nobody has opened in 60 days. The 2026 numbers back the gut feeling here. Roughly 30 percent of businesses report paying for redundant software, and a median AI spend of $2,200 a year spread across overlapping tools is money leaking with no receipt for the return.
Do not skip the free tier. ChatGPT, Claude, and Gemini all sit in most stacks now, often two of the three running at once for no reason anyone can explain. Free is not the same as harmless when three people paste customer data into three different chatbots with three different privacy settings. Write them all down.
While the rows are fresh, add a sixth column you can fill in 30 seconds each: the one job you bought the tool to do. Be specific. Not “marketing,” but “writes the weekly email.” Not “admin,” but “summarizes voicemails.” If you cannot finish that sentence for a paid tool, you found your first cut. Tools bought during a busy week, for a job you can no longer name, are the core of the sprawl problem. The point of the column is to force the honest answer out of your own notes instead of leaving it to memory.
Sort every tool into one of four boxes
Now make a decision on each row. Every tool lands in exactly one of four boxes. Keep, kill, consolidate, or connect. No fifth box, no “maybe later,” because maybe later is how you got here.
Keep is for a tool that does one clear job, gets used weekly, and you can name the time or money it saves. Kill is for anything nobody has touched in 60 days or anything whose job a tool you already pay for can do. Consolidate is for overlap. If you pay for a standalone AI writing app and you also pay for ChatGPT Plus, one of those is redundant. Pick the one your team actually opens and cancel the other. Connect is the smallest box and the most valuable. It holds the keepers that would be worth far more if they talked to each other.
Be ruthless in the kill box. A $29 a month tool you stopped using in March is $261 gone by year end. Three of those is most of your annual AI budget spent on nothing. The point of the audit is not to spend less on principle. It is to stop paying for tools that lost the job months ago.
Wire the survivors together so the work moves on its own
This is where the 14 percent number turns into your advantage. Most owners stop at “I have the tools.” Integration is what separates a pile of subscriptions from a system that does work while you sleep. The survey data is blunt about it. Three in four businesses that adopted AI have already hit a bad outcome from disconnected tools, usually the same data entered twice, or a lead that sat in one app while a follow-up tool waited in another.
Take a concrete example. Say a new lead fills out the form on your site. The job you want is: capture the lead, draft a personal reply, and log it in your CRM, with no human retyping anything. You can stand that up this week for under $30 a month. Zapier on a starter plan, around $20 monthly, watches the form. It passes the lead to an AI step that drafts a reply in your voice using a prompt you write once. Then it drops the contact into your CRM, whether that is HubSpot’s free tier or something you already pay for. One automation, three tools you mostly own already, and a job that used to eat fifteen minutes per lead now runs in seconds.
Start with one workflow, not ten. Pick the task you repeat most and hate most. For a lot of operators that is lead follow-up, invoice chasing, or turning a finished job into a review request. Build that one, watch it for a week, then build the next. A connected pair of tools you trust beats a drawer full of trials you never open.
Write down the result before you build, so you can tell later whether it worked. For the lead example, that is something you can check: every web form lead gets a drafted reply within two minutes, and lands in the CRM with no retyping. After a week, look at the log. If five leads came in and all five got handled, the automation earned its $20. If it broke on the third one, you learned that for $20 instead of finding out from an angry customer. Cheap tests beat big plans. The owners who win with AI this year are not the ones with the most tools. They are the ones who can point at a workflow and say what it does, what it cost, and what it saved.
What this buys you, and the signal to watch
The owners who run this audit tend to land in the same place. They cut one or two dead subscriptions, consolidate a redundant pair, and redirect the saved money toward connecting the two or three tools that earn their keep. The payoff shows up fast. Managers using AI well report saving north of seven hours a week, and most of that comes from the connective work, not from any single clever app.
The signal to watch over the next 30 days is consolidation moving inside the platforms you already pay for. The big assistants and the small-business suites are absorbing features that used to be separate paid products, from writing to scheduling to basic automation. When the tool you already own ships the feature you were about to buy, the right move is to cancel the trial and use what you have. Keep your audit spreadsheet open. Run it again at the end of the quarter. The business that knows exactly what its AI does, and what it costs, will outrun the one still collecting tools it never turned on.
