I thought I had an AI strategy. I had a pile of subscriptions and a strong opinion about which ones looked smart. Those are not the same thing, and the gap between them cost me most of a year.
If you run a small business and you say you have an AI strategy, there is a good chance you actually have a stack of disconnected tools with no single bottleneck named. A strategy names the one problem you are fixing this quarter. A pile names the tools you bought. The confusion owners feel when they ask “which AI tool should I use” is not a tool-selection problem. It is a no-bottleneck-named problem.
Why do I keep buying AI tools and still feel behind?
You keep buying AI tools and feel behind because you are solving for coverage instead of solving for one task. Around 54% of small business owners say the hard part is not access but evaluation: they know AI exists and cannot tell which option fits their actual need. So they hedge. They subscribe to four things, use pieces of two, and pay for the privilege of staying confused. The money goes out monthly. The one broken task in their business stays broken.
I watched owners on small-business forums describe this exactly. Several AI subscriptions. Hundreds of dollars a month. And when you ask them what problem it all solves, the answer gets vague fast. Not because they are careless. Because nobody told them the first move is to name the bottleneck, not to shop.
What did I find when I audited my own tools?
When I audited my own setup I found I had installed about 260 AI skills and 150 of them had never run once. I did the exact thing I warn founders about. I treated the pile as the plan. I counted installs as if volume were progress, and more than half of what I owned had never touched a real task. I wrote the whole embarrassing tally up in my post on installed versus used, because the number is the point.
Here is what that pile was really doing. It let me feel equipped. Every new tool was a small hit of “now I am ready,” and readiness is not the same as output. I never named the one thing I needed the tools to do, so I never noticed that most of them did nothing. A trenchcoat full of subscriptions looks like a strategy from a distance. Up close it is just weight.
What does a real AI strategy look like for a small business?
A real AI strategy is one named bottleneck you fix and verify, and nothing else until that one is done. Not a tool list. Not a subscription tier. You pick the single task that eats your week, you point one tool at it, and you check whether the output is actually good enough to use without you fixing it. If it is, you keep it. If it is not, you cut it. That is the whole discipline.
The move that finally worked for me was boring on purpose. I stopped asking “what can AI do” and started asking “what one task do I dread every week.” Then I fixed that, and only that, before buying anything else. I wrote about the year I lost before I figured that out in my post on wasting a year before fixing one boring task. The lesson held: one bottleneck beats ten tools.
The CEO of an AI company told Inc. that most companies start with the wrong AI strategy, and that the fix begins with naming one problem rather than shopping for capability (reported by Inc.). I read that and felt caught. I had done the thing at the scale of 260 skills.
How do I pick between the tools I already pay for?
Pick between your current tools by naming the bottleneck first, then keeping only the tool that moves it. Do it in this order:
- Name the one task. Write the sentence: “The thing that costs me the most time every week is ___.” If you cannot finish it, you are not ready to buy anything.
- List what you already pay for. Every AI subscription, with its monthly cost next to it. Most owners have never seen this on one line, and the total is the wake-up call.
- Match one tool to the one task. One. The others go on a kill list until the first task is verified solved.
- Verify the output. Did it produce work you would use without rewriting it? If no, the tool failed, not you.
Before I trust any automation to run on its own, I run a short set of checks so it cannot quietly produce garbage for a month. I keep that as the checklist I run before I trust an AI automation. Run something like it before you let a tool touch anything a customer sees.
What do I wish I had known a year ago?
I wish I had known that the number of AI tools you own tells you nothing about your AI strategy, and often predicts the opposite. The owners who are winning with AI did not buy the most. They named the fewest problems and finished them. I had 260 skills and 150 dead ones before I understood that counting installs was a way to avoid naming the one thing that was broken.
If you take one thing from this: stop shopping. Open the sentence “the thing that costs me the most time is ___” and do not buy another tool until you can fill in the blank and check the box. That reframe is the whole of my AI consulting work, and it is the cheapest fix most businesses will ever make, because it usually means canceling things, not adding them.
I was wrong for a year. The fix was not a better tool. It was one named problem and the nerve to ignore everything else until it was done.