12 Questions to Ask an AI Calling Vendor Before You Sign
A checklist for lenders comparing AI calling vendors: 12 questions on CRM, transfers, hours, caller ID, opt-outs, scripts, pilots and branch reports.
Before signing, ask an AI calling vendor what happens around the call, not just how the voice sounds. Does every outcome land in your CRM the same day? Do interested borrowers go live to their own branch? Who handles calling hours, caller ID and opt-outs? What will the agent never say? And what does success look like in numbers, by branch, before the pilot starts?
Ask about everything around the call, not the voice. The voice sounds good at every vendor now. What decides whether AI calling works at a lender is what happens before and after: where the outcome goes, who picks up when a borrower says yes, when calls go out, and how you'll know it worked.
Below are the 12 questions we'd ask if we were on your side of the table. Each one has why it matters and what a good answer sounds like.
Print it. Send it to the vendor before the next call. How they answer tells you more than any demo.
Why the demo won't tell you much
Here's how most vendor meetings go.
It's Thursday at 2 p.m. Your regional ops lead and someone from compliance are on the call. The vendor plays a recording. The voice pauses naturally. It laughs at the right moment. Everyone is impressed.
Nobody asks what happens when a borrower in a small Georgia town says "yes, I'm interested" at 4:55 p.m. and the branch closes at 5.
That's the problem. AI making a phone call is easy now. Making it work inside a branch network is the hard part, and a demo only shows the easy part. If you want the longer version of this argument, read what happens after the AI call.
So here are the questions, grouped by what they protect.
Questions about what happens after the call
1. Does every outcome land in our CRM the same day?
Why it matters. If call results live only in the vendor's dashboard, your branch managers won't see them, and nobody can follow up. The call becomes a dead end.
A good answer sounds like: "Every call writes back to your CRM record the same day: answered or not, interested or not, transferred or not, opted out or not. Your staff never have to log into our tool to act."
2. How does an interested borrower reach a person?
Why it matters. Interest fades in minutes. A borrower who says yes and then waits for a callback often isn't there when the branch calls. We compared both approaches in live transfer vs. callback.
A good answer sounds like: "They're transferred live to their own local branch while it's open. If the branch is closed or doesn't pick up, the borrower gets a scheduled callback and it's logged."
3. Who owns what happens after the transfer?
Why it matters. A transfer isn't a loan. Someone has to take the call, work the file, and record the result. If nobody owns that, the numbers look great and nothing funds.
A good answer sounds like: "Your branch owns the conversation after the transfer. We own getting the right borrower to the right branch, and we show you which transfers were picked up and which weren't."
4. How will we see results by branch?
Why it matters. A network-wide total hides the branch that never answers transfers. Your regional managers need to see their own branches.
A good answer sounds like: "A board by branch: calls placed, people reached, live transfers, callbacks. Updated daily, not in a monthly deck."
Questions about hours, numbers and opt-outs
5. When will you call, branch by branch?
Why it matters. Federal rules limit telemarketing calls to between 8 a.m. and 9 p.m. at the called person's local time, per 47 CFR 64.1200. But for live transfers the real limit is tighter: your branch has to be open.
A good answer sounds like: "We set calling windows per branch, inside each branch's own hours and time zone. No calls go out when nobody can take the transfer."
6. What number shows up on the borrower's phone?
Why it matters. People screen calls. A Pew Research Center survey from July 2020 found 80% of Americans generally don't answer calls from unknown numbers. A random out-of-state number makes that worse.
A good answer sounds like: "A local number per branch, registered with the carriers, with caller ID that says who's calling. If the borrower calls back, they reach something that makes sense."
7. How does an opt-out reach the same Do-Not-Call list everyone uses?
Why it matters. If a borrower tells the AI "don't call me," your branch staff shouldn't call them next week. The same federal rule, 47 CFR 64.1200, says a request to stop calls has to be honored within a reasonable time, no more than ten business days.
A good answer sounds like: "An opt-out on our call writes straight to your internal Do-Not-Call list, the one your branches and your mail vendor already use. We check that list before every call."
8. How is consent handled?
Why it matters. In February 2024 the FCC ruled in FCC 24-17 that AI-generated voices count as "artificial" voices under the TCPA. That means the consent rules for artificial-voice calls apply.
A good answer sounds like: "We'll show you what we check and how we record it. Your counsel decides what consent covers which list." Be careful with any vendor who says "don't worry, we're compliant." That's your counsel's call, not theirs. Our compliance checklist for lenders lists what to bring to that conversation.
This is educational, not legal advice. Review consent and disclosures with your own compliance counsel.
Questions about what the agent says
9. What will the agent never say?
Why it matters. One sentence like "you're approved" or "your rate will be" can create a problem that no call volume is worth.
A good answer sounds like: "It never promises approval, never quotes terms beyond your approved script, never makes a credit decision, and never pretends to be a person. When a borrower asks something outside the script, it offers to connect them with their branch."
10. Who reviews calls, and how does the script get better?
Why it matters. The first script is never the best one. Borrowers ask things you didn't expect. If nobody listens to calls, the same mistake repeats a thousand times.
A good answer sounds like: "A named person on our side reviews calls every week, flags problems, and proposes script changes. Nothing changes in the script until your team approves it."
Questions about the pilot
11. What does the pilot look like, and what does success mean in numbers?
Why it matters. This is where most pilots fail. They start without a finish line, so they never end. They just fade.
A good answer sounds like: "One branch, one list, a fixed number of days. Before the first call we agree on what we'll measure: calls placed, people reached, live transfers, and live transfers per 1,000 calls." We walk through this in how to run a 14-day pilot.
For a reference point, at Sunset Finance, a network of consumer lending branches in Georgia, the pilot produced about 8 live transfers for every 1,000 calls placed. Your list and offer will be different. The point is to have a number to compare against.
12. What happens if it doesn't work?
Why it matters. Every vendor is confident before you sign. The question is who carries the risk if the pilot flops.
A good answer sounds like: a specific answer, in writing. Not "we'll work with you." Something with a date and a condition.
When AI calling isn't the right choice
Honest answers to these questions will sometimes tell you not to buy.
If your list is a few hundred people, your branch staff can probably call it themselves over a couple of weeks.
If your borrowers need a long, sensitive conversation on the first call, an outsourced team of people may fit better. We compare the two in AI calling vs. a call center.
If you have no one at the branch to take transfers during the day, fix that first. An AI that transfers into an empty office makes things worse.
AI calling fits when you have thousands of people to reach, a simple first question to ask them, and branches that can pick up when someone says yes.
What waiting on the decision costs
You're probably thinking you'll take a few more months to compare vendors.
That's fair. Just know what keeps moving while you do.
Every month, the borrowers who paid off get a little further from the last time you talked to them. New applications keep arriving after 5 p.m. and waiting until the morning.
And your past customers aren't only yours. As the FTC explains, a lender can ask a credit bureau for a list of people whose credit reports meet its requirements and send them offers. The borrower who paid you off can land on someone else's list.
The questions above take one meeting to ask. For what it's worth, here's how we'd answer the last one: start with one branch, pay nothing up front, and only pay for results: a booked meeting or a reactivated customer. If it doesn't work, there's nothing to pay.
If you want to hear what your borrowers would hear, the AI will call you. Leave your number.
Questions lenders ask
What is the most important question to ask an AI calling vendor?
Whether every call outcome lands in your own CRM the same day. If it only lives in the vendor's dashboard, your branches can't act on it and you can't measure it.
Should the AI transfer calls live or schedule a callback?
For interested borrowers, a live transfer to their own branch while it's open is usually better, because interest fades fast. A callback makes sense after hours or when the borrower asks for a specific time.
What should an AI calling pilot measure?
Agree on it before the first call: calls placed, people reached, live transfers, and live transfers per 1,000 calls, all by branch.
Can the vendor tell us our calls are compliant?
No vendor should be your legal opinion. Ask how they handle consent, calling hours, caller ID and opt-outs, then review it with your own counsel.