Most small business owners buy automation once, from someone they found through a search or a referral, at the moment they know least about it. That is an uncomfortable position, and the industry is young enough that the range of competence is very wide.
You do not need to understand the technology to hire well. You need to ask the right questions and listen to the shape of the answers.
Here are the ones I would ask. They apply to me too, and I have written what I think a good answer sounds like.
About the price
1. Is this a fixed price or an estimate?
You want fixed, in writing, before work starts. If it is an estimate, ask specifically what would move it and by how much. "It depends on scope" is not an answer if the scope is already agreed.
Good answer: a number, in writing, with what is included and what is not.
2. What are the running costs, and will you show me the provider's invoice?
Every automation has ongoing costs: AI usage, text messages, phone minutes, hosting. Anyone saying it runs for free is either not thinking or not telling you.
Good answer: a realistic monthly figure, an offer to show you the actual bill, and the option to put the accounts in your own name.
3. What is explicitly not included?
Ask it in those words. The interesting part of any quote is its edges.
Good answer: a clear list. Revisions beyond a point, integrations not agreed, third-party costs, support after handover.
4. What would you tell me to buy instead of building this?
The most revealing question on the list. There is almost always an off-the-shelf product that does some of what you want, and a consultant who cannot name it either does not know the market or does not want you to.
Good answer: a specific product, an honest comparison, and a reason why building is or is not better for you. Sometimes that reason should be "it is not, buy the product."
About what happens when it goes wrong
5. What happens when the AI gets something wrong?
Not if. Every one of these systems produces confident wrong answers sometimes. The question is what has been built to contain that.
Good answer: it works from your real data rather than its memory, it drafts for approval rather than acting alone where the stakes are real, it says it does not know rather than guessing, and it escalates to a human.
6. Can I switch it off myself?
You should not need to phone anyone to stop it at 9pm on a Saturday.
Good answer: yes, here is how, and here is where you can see what it has been doing.
7. What happens if it breaks in six months?
Providers change things. Systems get updated. Something that worked in January stops working in June.
Good answer: a clear support arrangement with a price on it, or an honest "you would call me and I would quote it". Both are fine. Vagueness is not.
About ownership
8. Do I own the code and the data?
Good answer: yes, and here is how you would take it elsewhere.
If the answer is no, that may still be acceptable, but you are renting, not buying, and the price should reflect that. Know which one you are doing.
9. Can I export everything?
Customer records especially. If your data lives somewhere you cannot get it out of, you have a problem you will discover at the worst moment.
10. Is my business data used to train anything?
Good answer: no, with the provider's written policy offered rather than just asserted.
About them
11. What have you actually built, and can I see it?
Working software, not slides. This industry has a lot of people who are three months ahead of their clients and describing it as a decade of experience.
Good answer: something you can look at or ring. For what it is worth, mine is the receptionist on (218) 818-1023 and the chat panel on this site, and I say plainly on the AI tools page that my AI work so far is my own products rather than client builds.
12. When would you tell me not to do this?
The single best question here. Anyone who cannot name a situation where you should walk away is selling rather than advising.
Good answer: several specific situations. Volume too low to pay back the build. A process still changing every month. A job where being wrong is expensive. Wanting it to replace a person.
Three answers that should worry you
"We can guarantee you a return on investment." Nobody can know that before seeing your books. Hours removed and running costs are measurable. Revenue is not a promise anyone can make honestly.
"We can get you to number one on Google." Rankings depend on competition, reviews, links, content and time. Anyone guaranteeing position one is guaranteeing something outside their control.
"You will not need to check it." You will, at least at first. Anyone saying otherwise has either not built one or is not telling you how it behaves in month two.
What a good first conversation feels like
It should be mostly them asking about your week rather than describing their technology. It should produce specific opportunities with numbers attached rather than possibilities. It should include at least one thing they would not automate. And it should be free, or cheap enough that walking away costs you nothing.
If you leave the call knowing less about your own business than when you started, that was not a good conversation regardless of how impressive the demo was.
Ask me these
Everything above applies to hiring me, and I would rather you asked. My prices are published instead of quoted on a call, the audit is free, and it ends with "not yet" often enough that it is mentioned on every page of this site.
If you would rather see what the work looks like before talking to anyone, the contractor hub lays out the specific builds and what each one costs.