Most articles on this topic answer this question with a thinly disguised pitch. We're going to do something different — we're going to give you the test we run before we take an engagement. If a business doesn't pass it, we say so, and we don't take the work. Not because we're being noble; because we've watched what happens when AI is sold into the wrong situation, and the spreadsheet six months later is ugly.
Here are the four questions.
Question 1: Is there a measurable leak?
AI doesn't create new revenue out of thin air. It plugs leaks. So before anything else: where is the leak, and how big is it?
The most common leaks in a local service business:
- Missed calls — if 20% of inbound calls hit voicemail and 60% of those don't call back, you can put a number on it: (calls/month) × 0.20 × 0.60 × (avg job value) × (close rate).
- Slow follow-up — quotes that don't get a response within 48 hours close at half the rate of quotes that do. If you send 50 quotes/month and 20 sit cold, that's a measurable leak.
- Missing reviews — every Google review you don't ask for is a customer you might not have asked for either. Businesses with 100+ reviews close roughly 2x the rate of those with 10.
- Bad scheduling — gaps in the calendar that wouldn't exist if rebooking was systematic.
If you can't put numbers on a leak, AI isn't worth it for you yet. Not because AI couldn't help — because you don't have the data to know whether it did. Spend two weeks measuring before you spend any money.
The minimum viable measurement
For two weeks, write down: how many inbound calls came in, how many you answered, how many quotes you sent, how many converted, how long until you followed up. That's the before. Without it, "AI is worth it" is just a feeling, not a fact.
Question 2: Is the leak at least 3x the cost of fixing it?
This is where most owners miss the math.
If your missed-call leak is $1,500/month and an AI setup costs $1,500/month, the math doesn't work even when everything goes right. You break even. Best case. And reality is messier than best case — you'll lose 1-2 weeks during onboarding, you'll have edge cases that take time to debug, customers will call about something the AI didn't handle.
For the math to be reliably positive, the leak should be at least 3x the cost of fixing it. That gives you margin for the messiness.
Worked example for a typical service business at $1,500/month AI setup:
- Leak: 25 missed calls/month, 60% don't call back, $400 average job, 50% close rate
- Recoverable: 25 × 0.60 × $400 × 0.50 = $3,000/month
- Ratio: $3,000 / $1,500 = 2x — borderline. We'd push harder on this one.
Now scale it up:
- 50 missed calls/month, 60% don't call back, $600 average job, 50% close rate
- Recoverable: 50 × 0.60 × $600 × 0.50 = $9,000/month
- Ratio: 6x. Clear yes.
And down:
- 10 missed calls/month, 50% don't call back, $200 average job, 30% close rate
- Recoverable: 10 × 0.50 × $200 × 0.30 = $300/month
- Ratio: 0.2x. Clear no — at this scale, a $50/month off-the-shelf missed-call texter is the right answer, not a managed setup.
Question 3: Will you actually use the data the AI gives you?
Half the value of an AI setup isn't the agent. It's the data the agent surfaces — what gets asked most, what objections come up, where conversations go cold, which job types convert best.
That data is gold if you read it. It is wallpaper if you don't.
The honest test: can you commit 30 minutes a week to reading the digest? Not "I'll try" — actually commit. If the answer is no, AI is going to do its thing and you're going to keep running the business the same way, and you'll never know if it's working. That's a $1,500/month problem.
The owners who get the most out of AI aren't the ones who automate the most. They're the ones who pay attention to what the automation reveals.
Question 4: Are you willing to give it 90 days?
This is the question that separates owners who succeed with AI from owners who don't.
Month 1 is messy. The agent gets things wrong. You'll have to listen to call recordings and tell us what to fix. There will be a moment around week 3 where you'll wonder if this was worth it.
Month 2 is when it starts to feel right. The agent has been tuned for your specific objections, your specific tone, your specific schedule. The reports start to make sense.
Month 3 is when the math becomes visible. The recovered jobs add up, the review count climbs, the quotes-cold rate drops. You can put a number on what changed.
If you're not willing to ride out the messy first 30 days, AI isn't going to work for you. Not because AI is bad — because the work of teaching a system your business takes time, and shortcutting it produces an agent that's mediocre forever.
The "is AI worth it" scorecard
| Question | Pass |
|---|---|
| Is there a measurable leak? | You can put real numbers on it |
| Is the leak ≥ 3x the cost? | Recoverable ≥ 3 × monthly fee |
| Will you read the data? | 30 min/week commitment |
| Will you give it 90 days? | Yes, including the messy month 1 |
Four passes: AI is worth it. Go.
Three passes: probably worth it; the missing one tells you what to fix first.
Two or fewer: it's not worth it yet. Fix the gaps and revisit.
When the answer is no — and we'll tell you
Here are the situations where, in our experience, AI isn't worth it for a small business — at least not yet.
You're under $20K/month in revenue
A $1,500/month tool against a $20K business is 7.5% of revenue. The leaks at that scale typically aren't large enough to clear the 3x bar. Better answers: a $50/month missed-call texter, a Calendly link in your email signature, a one-page Notion script for follow-ups. Revisit AI at $40K+/month.
Your work is high-judgement, low-volume
Custom builds, complex consulting, anything where every client conversation is a one-off and the customer is paying for your specific brain. AI doesn't help here. A part-time person who learns your style helps. Don't try to automate what is fundamentally bespoke.
You can't articulate what good looks like
"I want AI to help with my business" is too vague to act on. If you can't finish the sentence "I want AI to handle ___ so that ___," you're not ready. Spend a month watching where the time goes. Then come back.
You hate technology and won't engage with the dashboard
It's fine to dislike software. But AI requires you to read its work for the first 60 days. If the dashboard is going to sit unread and the call recordings are going to go unsampled, the agent will drift, and you'll be paying for something that's quietly getting worse. Honest signal: AI isn't a fit if you won't look at it.
You're trying to fix a hiring problem
If your real problem is "my receptionist is bad," replacing them with AI typically doesn't fix it — it just changes the failure mode. Fix the hiring problem first; then layer AI on top of a working setup.
How we approach this with new clients
The first conversation we have isn't about features. It's about whether you pass the four questions above. About a third of the time, we end the conversation with "you don't need us yet" — and we say what to fix first. We'd rather give that answer up front than take money for an engagement that won't pay back.
That posture is built into how we price. We charge a fraction of the value the system creates, and we stay with you while it compounds. If we can't see a path to that ratio, we shouldn't be the people you hire.
The summary
AI is worth it for a small business when there's a measurable leak, the leak is at least 3x the cost of fixing it, you'll read the data, and you'll give it 90 days. If those four are true, the math reliably works. If they aren't, no amount of vendor enthusiasm will make it work.
If you want a specific answer for your business, the playbook walks through the leaks; reply to the email it triggers and we'll send back a 2-paragraph read on whether you pass the test — and what to fix first if you don't.