How to Choose an AI Appointment Setter: A Buyer's Checklist

Every AI appointment setter demos the same way. A clean lead comes in, the agent answers in a friendly voice, asks two questions and books a slot. Nobody demos the lead who asks something off-script at 11 p.m., the booking that never reached the CRM, or the agent that still mispronounces your company name in week six.

Choose an AI appointment setter on seven things: response speed, what it does when stuck, how it captures consent, which systems it writes to, whether you can read every transcript, who tunes it after launch, and what the price covers. Most are invisible in a demo, so ask directly.

Here is what to ask, why each question matters, and what a good answer sounds like.

1. How fast does it answer, measured from what?

Speed is the one thing an AI setter should beat a person at, so it is worth pinning down.

Ask for the number and the starting point. "Instant" can mean the agent starts dialing instantly, or it can mean a workflow fires in 30 seconds, a queue picks it up, and the call connects two minutes later. Those are different products.

The reason to care: in an audit of 2,241 US companies, Harvard Business Review found only 37% responded to a web lead within an hour, and firms that reached a lead within an hour were nearly seven times as likely to qualify it as firms that waited an hour longer. More on the research in our lead response time breakdown.

A good answer gives you a number you can test. At Executive Stride, agents built on StrideOps respond in under 2 seconds, and you can watch that on a test form fill before you sign anything.

Ask for: a live test. Fill out your own form while the vendor watches, and time it with your phone.

2. What happens when it cannot answer?

This is the question that separates the products. Every agent handles the easy path. The question is the design of the failure.

There are three behaviors an agent can have when a question falls outside what it knows:

• It guesses. Worst case. In a regulated industry it is a liability, and in any industry it burns the lead.
• It deflects and keeps pushing for the booking. Common, and it reads as a script. Leads notice.
• It says it does not know and offers a person. This is the one you want.

Ask how a hand-off actually works. Does it warm-transfer to a live phone, text your team, create a task, or just leave a note nobody reads? Ask what happens at 11 p.m. when there is no person to transfer to, because "transfers to a rep" means nothing outside business hours.

Ask for: a recording or transcript of a conversation where the agent got stuck. If the vendor cannot produce one, they have not looked.

3. How is consent captured, and by whom?

Consent is not a feature of the agent. It lives on your forms, in your records and in your opt-out handling, and it is settled before the first message goes out.

On February 8, 2024, the FCC issued a Declaratory Ruling (FCC 24-17) confirming that AI-generated voices count as "artificial" voices under the Telephone Consumer Protection Act, so calls using them need the same prior express consent as other artificial or prerecorded calls. For telemarketing that means prior express written consent.

Two things to pin down:

• Who writes the disclosure on your form, and will they look at the forms you already have. A vendor who never asks to see your lead forms is not thinking about this.
• What consent records you keep. The form version, disclosure text, timestamp, source page and number. You want those in your system, not only in the vendor's.

Also ask what the agent will not do. Ours only contact people who asked to hear from you. Purchased lists and scraped numbers are outreach we do not run, and a vendor who offers to warm up a cold list with AI voice is selling you a problem. Our SMS terms show the language that governs our texts.

This is not legal advice. Confirm your setup with your own counsel.

Ask for: the exact disclosure wording they would put on your form, and where consent records are stored.

4. Which of your systems does it write to?

An appointment that exists only in the vendor's dashboard is not booked. Ask the question as a list, not as "do you integrate?"

• Does it write the appointment to your calendar, with the right owner, duration and buffer?
• Does it create or update the contact and opportunity in your CRM, or just log a note?
• Does it respect existing availability rules, like round robin across three loan officers or a travel radius?
• Does it push the outcome back to your ad platforms so you can optimize on booked calls rather than form fills?
• If the lead reschedules by text, does the calendar move?

Native integrations are cheaper and more reliable than custom ones. Ask which category yours falls into, because that is usually the difference between going live in two weeks and going live in two months.

Ask for: a screenshot of a booked appointment inside the CRM you already use.

5. Can you read every conversation?

You should be able to open any call or text thread and read it end to end, with the recording, without asking anyone.

That matters for three reasons. It is how you find out why good leads are not booking. It is how you catch the agent saying something it should not. And it is your record if a lead ever disputes consent or what was promised.

Ask whether transcripts are searchable, how long they are retained, whether you can export them, and who else can see them. Then ask about call recording notices. Federal law allows recording with one party's consent (18 U.S.C. 2511), but a number of states require all parties to consent, so the agent needs to announce recording where that applies.

Ask for: access to a transcript view during the sales process, not after.

6. Who tunes it after launch, and how often?

An agent is not finished when it goes live. It gets better from real conversations: the question nobody anticipated, the objection that keeps killing bookings, the name it keeps mangling.

So ask who does that work. The honest answers are "you do," "we do for the first 30 days," or "we do continuously." All three are workable. What matters is that you know which one you bought.

Then ask what the loop looks like. Someone reviews conversations, finds the pattern, updates the knowledge base or the script, and tests the change before it ships. If the answer is that the model learns automatically, ask to see what changed last month and who approved it.

Our agents are tested in simulated calls before they touch a real lead, and full optimization takes 30 to 60 days of real conversations. Expect any agent to get sharper over its first two months.

Ask for: the last three changes they made to a live client's agent and why.

7. What does the price actually cover?

Pricing models differ enough that monthly numbers do not compare. Convert everything to cost per booked appointment and cost per show, which is the method in our AI appointment setter cost breakdown.

The four shapes you will see:

Then ask the questions that reveal the rest of the number. Is A2P 10DLC registration included? US business texting to local numbers requires it, and there are carrier fees and approval delays, which our A2P 10DLC guide walks through. Are phone numbers yours or theirs? Is script writing included or billed? What is the term, and what happens to your numbers and transcripts if you leave?

Ask for: a quote you can divide by expected bookings. If it cannot be turned into cost per show, keep asking.

Red flags

A few answers should slow you down:

• A guaranteed number of appointments before anyone has seen your leads, offer or market.
• No consent conversation. If nobody asks about your forms, nobody is managing your risk.
• Willingness to call cold lists. It tells you how they treat compliance everywhere else.
• No transcript access, or access only through a support request.
• "The AI handles everything." No agent handles everything. The good ones hand off well.
• A demo you cannot interrupt. Ask to talk to the agent yourself and go off-script.

Should you buy one at all, or build it?

Three honest paths:

Build it yourself. If you enjoy this work and have the hours, the tooling is good enough now that a competent operator can get a decent agent running. The free AI Operator Kit is a reasonable place to see what the build actually involves before you spend anything.

Hire a human setter. Better at upset callers and genuinely complex conversations, limited to the hours you pay for. Our AI vs human appointment setter comparison covers the split, and how much to pay an appointment setter covers the pay structures.

Buy it done for you. You pay more in cash and less in time, and the seven questions above are how you tell the vendors apart. Our done-for-you AI appointment setters page covers what a full build includes.

Many teams end up with the AI on first response and a person on the hard conversations, which is the arrangement we build most often.

A one-page scorecard

Score each vendor out of 2 and compare the totals rather than the monthly prices.

Anything under 9 out of 14 is a vendor you will be managing rather than one managing your leads.

FAQ

What is the best AI appointment setter?

There is no single best one, because the right choice depends on your lead volume, which CRM you run and how regulated your industry is. Compare on response latency, hand-off behavior, consent handling, integrations, transcript access, who tunes it and cost per show.

What should I ask an AI appointment setter vendor?

Ask for a timed live test from your own form, a transcript where the agent got stuck, the exact consent wording they would add to your forms, a screenshot of a booking inside your CRM, and the last three changes they made to a live client's agent.

How long does an AI appointment setter take to set up?

A straightforward build on native integrations takes a few weeks. Custom integrations, multiple scripts and A2P 10DLC registration add time. We get most clients live in 2 to 3 weeks.

Is there an AI appointment setter that works for my industry?

Most work across industries, but regulated ones need more care. Mortgage, insurance and lending each have disclosure and licensing rules that shape what the agent is allowed to say. See mortgage AI compliance for how that plays out in one of them.

What is an AI setter?

An AI setter is a voice or SMS agent that answers inbound leads, asks your qualifying questions and books the ones that fit onto your calendar. It is the first-response layer, not a closer.

Run the seven questions on us

If you are comparing options, bring this list to a call with us and ask all seven. We will show you a timed test, a stuck transcript and a booking inside your CRM. Book a Stride Scaling Call.