Most small businesses aren’t asking whether AI works anymore. They’re asking whether it works for them, this year, without turning into another subscription nobody opens. That’s a fair question, and it has a better answer than “it depends.”
Here are ten questions we run through before any automation or AI build. Answer them honestly. Yes or no. No partial credit.
Why a checklist at all
Because the failures are predictable. When AI projects stall in a small business, it’s rarely the technology. It’s a fuzzy goal, or messy data, or nobody assigned to check the output. Those problems show up the same way in a plumbing company, a church office and a law firm.
We’ve run 1,258 projects in seven months across 43 active clients. When something stalled, the cause almost always traced back to one of the ten questions below.
The 10-point check
1. You can name the task that eats your week
Not “marketing” or “admin.” A specific task. Chasing hosting renewals. Building the monthly report. Answering the same six emails about hours and pricing. If you can’t name it in one sentence, AI can’t do it yet, because nobody knows what “it” is.
2. The task repeats
AI earns its keep on work that happens every week or every month. A one-time project might still be worth doing with AI, but the real payoff is the job that comes back. Our invoice automation matters because renewals never stop coming.
3. The inputs already live somewhere digital
Email. A spreadsheet. Your invoicing system. A shared drive. If the information starts on a clipboard in a truck, step one is getting it into something a system can read. That’s not a dealbreaker. It’s just the first job.
4. You know what “done right” looks like
Could you write down what a good result looks like? A correct invoice. A report with these five numbers on page one. A reply that sounds like you and not like a bank. If the standard only lives in your gut, write it down first. AI follows a clear standard very well. It can’t read your mind.
5. Someone owns the review
This one is non-negotiable. Every output needs a human who looks at it before it counts. In our studio, AI does the work and a human principal signs off on every piece. Your business needs the same thing, even if review takes five minutes a week. No reviewer means no project.
6. Your data is reasonably clean
Not perfect. Reasonably clean. Customer names spelled one way. One spreadsheet, not four competing versions. Dates that are actually dates. Garbage in, confident garbage out.
7. You know what should never leave the building
Client files. Medical details. Anything under a confidentiality obligation. Before any AI tool touches your work, decide what’s off limits and which tools are allowed near what. For law firms and healthcare offices this is the first question, not the seventh. We walk through it in Should Your Law Firm Use AI?
8. Your accounts are actually yours
Your domain, your Google or Microsoft account, your invoicing login, your social pages. They should be in the business’s name with credentials you control. A common story: a relative set up the Facebook page years ago and now nobody can get in. Automation can’t connect to accounts nobody can access.
9. You’re fine with a draft that needs a tweak
AI gets you most of the way. The last stretch is you reading it, fixing a line, approving it. If your expectation is zero touch, forever, you’ll be disappointed. If your expectation is a lot less work on a job you hate, you’ll be happy.
10. You’ll know if it worked
Pick the measure before you start. Hours back each week. Invoices that go out on time. Follow-ups that stop falling through the cracks. Without a yardstick, every AI project turns into a vibe, and vibes don’t justify a monthly bill.
How to read your score
Count your yeses.
Eight to ten. You’re ready. Pick the task from question one and build it. The hard groundwork is already done.
Five to seven. You’re close. Look at which questions you missed. Usually it’s data cleanup (6), account access (8) or nobody owning review (5). Those are fixable, and fixing them is useful even if you never touch AI.
Four or fewer. Don’t buy anything yet. Seriously. Start with questions one and four. Name the task and define “done right.” That exercise alone tends to show you whether the problem needs AI, a better process, or just a phone call to the right vendor.
What we’d add as question eleven
Would you rather own the system or rent the advice? Some businesses want a consultant to tell them what to do so their own people can build it. Others want the whole thing built, running and handed over. Neither is wrong, and it changes who you should hire. We laid out the difference in Hiring an AI Consultant vs. Hiring a Studio That Runs on AI.
The honest part
A lot of businesses that take this check find out the first win isn’t AI at all. It’s cleaning up a spreadsheet, or getting the Facebook login back, or writing down how a job should go. That’s fine. That work makes every future tool easier to use.
And when the boxes are checked, AI tends to deliver. Not in a flashy way. In a quiet, the-invoices-went-out-and-nobody-had-to-think-about-it way. That’s the kind we like. We wrote up a real example in The Invoice That Sends Itself.
If you want to see how we take a business from this checklist to a running system, How It Works walks through the steps.
Scored yourself and want a second opinion on the result? Start here.