Probably less than you think. But probably more than zero.
Most of the advice about AI for small businesses starts with the tools: which chatbot, which assistant, which app. That's backwards. The useful question is where your business is losing customers and hours today, and whether AI is the cheapest way to stop it.
Count four things for one week
You don't need software for this. A notepad by the phone is enough.
- Calls or messages that waited more than an hour for a reply. Include voicemails and weekend messages.
- New inquiries that never got a follow-up. The quote you sent once, the form nobody answered.
- Hours your team spent typing the same details into different screens. Invoices into accounts, forms into your booking system, notes into your CRM.
- Appointments that no-showed without a reminder.
If all four are close to zero, you don't need AI yet. Keep doing what you're doing.
If one of them isn't, that's where you're losing business, and that's the one place AI should start. Not everywhere. Just there.
Why the first reply matters so much
People asking for a quote usually ask two or three businesses at once. In published response-time tests, real estate offices took a median of 39 minutes to reply to a new online inquiry, and law firms took a median of 13 minutes (Focus Digital summary). Most people don't wait that long. They book whoever answers first.
That's why answering every call and replying to every new inquiry within a minute is often the first job worth handing to AI. It's the leak you can't see, because the customer who went elsewhere never shows up in your numbers.
Put a number on it
Once you have a week of counts, two sums tell you what the leaks cost each month:
- Revenue at risk = missed or late inquiries per week × 4.3 × the share you think would have booked × what a new customer is worth.
- Admin cost = admin hours per week × 4.3 × what an hour of that time costs you.
A clinic that misses 10 inquiries a week, thinks 1 in 5 would have booked, and values a new patient at $300 has about $2,580 a month at risk. Add 8 hours a week of admin at $25 an hour, about $860 a month, and the total is roughly $3,440 a month.
Those are made-up numbers to show the arithmetic. Yours are the only ones that matter. The three-minute check does the sums for you, in your browser, without sending your numbers anywhere.
Four honest answers
| Your leaks each month | What that usually means |
|---|---|
| Under about $500 | Not yet. Fix the basics: a booking link on every page, one shared inbox. |
| About $500 to $5,000 | A little. One automation on the biggest leak, measured for a month. |
| About $5,000 to $20,000 | A fair amount. Several leaks. An audit puts them in order. |
| Over about $20,000 | A lot. Worth looking at a system built around how you work. |
The cut-offs come from simple payback. If one automation costs a few thousand dollars and fixes only half of a small leak, it takes more than a year to pay for itself, and there are better uses of the money.
Then grow it one job at a time
Whatever your answer, the order is the same.
- Write down how your business works. Prices, policies, and the answers your best person gives ten times a week. Think of it as your company brain. AI works from it, and new staff learn from it even before any AI is involved.
- Hand over one job in one part of the business: the leak that costs most.
- Measure it for a month. Keep it, fix it or stop.
- Then the next job. Each step is small, so each investment is too, and you only take the next one when the last one has paid.
That is how businesses end up running on AI without betting the business on it. For the bigger picture, see how businesses rebuild their work around AI, in plain words.
Keep a person in charge
Whatever you automate, one rule keeps it safe: reversible things can happen on their own, and irreversible things wait for a person. A reminder can go out automatically. A refund, a quote or a first message to a new customer should wait for approval until you have watched the AI do it well.
The question to ask instead
So the real question isn't "should we use AI?" It's "are we using technology to win business, or quietly losing it?"
Count for one week. Then you'll know.
Frequently asked questions
How do I know if my business needs AI?
Count four things for one week: replies that waited more than an hour, inquiries nobody followed up, hours spent retyping details, and no-shows without a reminder. If all four are close to zero, you don't need AI yet.
What should a small business hand to AI first?
The leak that costs the most each month. For many businesses that means answering every call and new inquiry within a minute, because people asking for a quote usually book whoever answers first.
Is it worth automating a small leak?
Usually not. If an automation costs a few thousand dollars and fixes only half of a small leak, it takes more than a year to pay for itself. Fix the basics first, such as a booking link and one shared inbox.
Should AI send messages to customers on its own?
Only things that can be undone, such as a reminder. Refunds, quotes and first messages to new customers should wait for a person to approve them until you have watched the AI do them well.
