Field notes
Twelve questions to ask an AI calling vendor before you sign
Every one of these has come up in a real conversation. Most have an answer a good vendor gives immediately and a bad one talks around. The dodge is the information.
Print it. Take it into the demo.
On consent and the list
1. Will you call a list I bought?
The answer you want is no. A vendor who will auto-dial a purchased list will auto-dial anything, and the exposure sits with you, not them.
2. What do you need to see before you’ll load my contacts?
They should ask about provenance — where the contacts came from, what the person agreed to, how old it is. If they only ask for the file format, nobody is thinking about the part that gets you sued.
3. Who is responsible if a complaint comes in?
You are. That is the correct answer and an honest vendor says it plainly. What you are testing is whether they will pretend otherwise to close the deal. A vendor who says they handle compliance for you is either confused or selling.
On the phone system
4. Whose EIN is the messaging registered under?
Yours. If it is theirs, you are on a shared sending brand, inheriting other clients’ reputation and unable to take yours with you when you go.
5. If I leave, what comes with me?
Numbers, registration, recordings, and your contact data. If any of those stay behind, that is a switching cost they built on purpose.
6. Whose account holds the phone numbers?
There is a defensible version of either answer, but you should know which it is before you find out during a dispute.
On what the thing actually does
7. Can I call it right now?
Yes, and here is the number. Any vendor selling a voice agent should have a live line you can dial without a form. If the only way to hear it is a scheduled demo where they drive, you are being shown a rehearsal.
8. What happens when it doesn’t know the answer?
You want a real mechanism — hand off, take a message, escalate. What you do not want is an agent that improvises. Confident invention is worse than “let me get someone who knows.”
9. What does it do when someone asks to stop?
Opt-out should be built into the message path, honored immediately, and permanent. Ask them to describe it, not to confirm it exists.
On the install
10. What happens between signing and going live, and how long does it take?
You are looking for stages and a range they will commit to. “Right away” means nothing was tested. A vendor who cannot describe their own process has not run it enough times to have one.
11. Do you test it before it takes a real customer call?
There should be a defined protocol that has to pass before live traffic. Ask what it covers. Calendar, handoff, consent path, voice, texts. If testing means “we tried it a few times and it seemed good,” the first real customer is the test.
12. When does billing start?
Setup and monthly should be separate, and you should know exactly what triggers each. Ours starts at go-live, not when the build opens, because charging a monthly fee for a system nobody can call yet is charging for nothing.
The pattern
Read back through those. Almost every one is asking the same underlying question: when this goes wrong, who is holding it?
Not because it will definitely go wrong. Because the answer tells you what kind of company you are dealing with, and you learn more from how a vendor handles that question than from any demo.
The good ones answer fast and specifically, including the parts that are not flattering — yes, the obligation is yours; yes, we will decline a list; yes, this takes weeks not days. The bad ones get vague exactly where the money is.
You do not need to be technical to run this list. You need to notice which questions produce a straight answer and which produce a paragraph.
This is not legal advice. It is a buyer’s checklist based on how we build installs. Consent and messaging obligations belong to the business sending the message, and you should have your own attorney review your program.