AI cold calling means using artificial intelligence to place sales calls, or to support the person who places them. In the US, an AI voice that cold calls a mobile number for telemarketing needs the prospect's prior express written consent. So for B2B outbound, the setup I trust is simple. A human does the talking. AI does the research, notes, follow-up and CRM.
That is how my own team runs it. My assistant has booked me 3 meetings in one hour of calls. She calls and talks. AI handles everything around the call: the summary, the follow-up email, the recap and the CRM update. This guide explains the two ways to use AI on cold calls, what the law says, what it costs and where each one fits.
- There are two models: an AI voice agent that makes the call, or a human caller with AI doing the work before and after.
- Since the FCC ruling of February 8, 2024, an AI-generated voice counts as an artificial voice under the Telephone Consumer Protection Act (TCPA).
- Telemarketing calls with an artificial voice to US mobile numbers need prior express written consent. A cold call has no prior consent, so an AI voice cannot legally make it to a mobile.
- Statutory damages are $500 per violation, or the actual loss if higher, and up to three times that when the violation is willful.
- Voice agent platforms publish prices from about $0.07 to $0.31 per minute. The legal exposure is the bigger number.
in
“My assistant books me 3 meetings in 1 hour. She calls and talks. AI orchestrates her ops:”
What Is AI Cold Calling?
AI cold calling covers two very different things. In the first, software dials the prospect and an AI voice holds the conversation. In the second, a person makes the call and AI prepares, listens and follows up. Most vendor pages mix the two. The difference decides your legal risk, your cost and how buyers react.
| Criterion | AI voice agent | Human caller with AI support |
|---|---|---|
| Who speaks | Synthetic voice driven by a language model | A trained person |
| What AI does | Dials, talks, qualifies, books or routes | Research, call notes, summary, follow-up email, CRM update |
| US law for cold calls to mobiles | Artificial voice: prior express written consent required for telemarketing | Live human voice: the artificial voice rule does not apply. Autodialer, Do Not Call and state rules can. A cloned voice counts as artificial |
| Cost you can see | Per minute, published from about $0.07 to $0.31 (Retell AI) | Caller time plus AI tools |
| Best fit | Calls to people who asked to be called and consented | First contact with B2B buyers who never opted in |
The second model is the one most sales teams can use for true cold outreach today. It is also where the time savings are real. A caller loses more time to notes, emails and CRM fields than to dialing. AI removes that work without putting a synthetic voice in front of a buyer who never agreed to hear it.
How Does AI Cold Calling Work?
AI cold calling works in five steps around the call: prepare, talk, summarize, follow up and record. For a human-led team, AI carries the steps before and after the conversation. An AI voice agent also takes over the talking, using speech recognition, a language model and text-to-speech.
- Before the call: AI researches the account and the person, checks a trigger event and drafts a one-line opener.
- During the call: the caller talks. The call is recorded and transcribed, where the law and the prospect allow it.
- Right after: AI writes a short summary with pains, objections and next step.
- Follow-up: interested buyers get the information they asked for by email, written by AI.
- Record: the CRM updates itself after key events, so the pipeline stays true without typing.
The after-call part is its own workflow. I break it down step by step in sales process automation after every call.
Is AI Cold Calling Legal?
AI cold calling is legal in the US when the AI voice calls people who gave the right consent. It becomes risky when an AI voice calls strangers. On February 8, 2024, the Federal Communications Commission announced a unanimous ruling: calls made with AI-generated voices are artificial under the TCPA. It took effect immediately.
Two rules follow from that. The TCPA bans calls with an artificial voice to mobile numbers without the called party's prior express consent. The FCC's rules go further for sales calls: telemarketing with an artificial voice to a mobile number needs prior express written consent. A cold call has no prior consent, so an AI voice cannot legally make it to a mobile.
B2B teams often assume business numbers are exempt. The residential line rule does not cover a company switchboard. The mobile rule has no business exception, and mobiles are exactly what outbound data tools now sell. In our data test, the two best providers found a mobile number for 68% and 70% of a 50-lead sample.
The cost of getting this wrong is set by the statute. A person can sue for $500 per violation, and a court can triple it to $1,500 when the violation is willful or knowing. Run the math on 1,000 calls to mobiles without consent: $500,000 to $1.5 million of exposure. This is general information, so ask your lawyer before you launch.
Consent is not the only rule. An artificial voice call must name the business at the start, and telemarketing calls need an automated way to opt out. Sales calls to consumers are limited to 8 a.m. to 9 p.m. local time. Some states add stricter rules, such as Florida's Telephone Solicitation Act.
How We Use AI on Cold Calls
The best use of AI on cold calls is everything except the conversation. We keep one rule: humans build the relationship and close, AI does everything else. My assistant makes the calls in her own voice. The four jobs below run on AI. That split is how she has booked me 3 meetings in a single hour of calls: she spent the hour talking to buyers and none of it on admin.
| Step | Who does it | What the AI output looks like |
|---|---|---|
| Call | The caller, live | Nothing. The conversation is hers |
| Summary | AI, after every call | What the buyer said, the pain, the objection, the agreed next step |
| Info for interested buyers | AI | A short email with our information, sent to buyers who showed interest |
| Meeting recap | AI | A recap email that confirms time, attendees and agenda |
| CRM | AI, after key events | Stage, notes and next task updated without manual typing |
The quality of the call still depends on the caller. A good opener and good questions matter more than any tool. AI makes her faster and keeps the record clean. It does not make a weak pitch strong.
If you want the questions that make those calls work, use my list of sales discovery questions.
What Does AI Cold Calling Cost?
An AI voice agent costs cents per minute. Retell AI publishes an all-in range of about $0.07 to $0.31 per minute, built from voice infrastructure at $0.055, text-to-speech, the language model and telephony from $0.015. At that rate, 1,000 five-minute calls cost roughly $350 to $1,550 before setup and testing.
The human-plus-AI model costs the caller's time plus the AI layer. The AI layer is small next to a salary. The bigger saving is time: every minute not spent on notes and CRM fields goes back into conversations. For an AI voice agent, the price per minute is the small number. The consent risk above is the one to budget for.
When Does an AI Voice Agent Make Sense?
An AI voice agent makes sense when the person expects the call and has consented to it. That covers a large set of useful calls. It does not cover first contact with a buyer who never heard of you.
- Speed to lead: calling back someone who filled in a form and ticked a clear consent box for calls.
- Qualification of inbound demand: a few questions before the call is routed to a rep.
- Reminders and confirmations for meetings already booked.
- Re-engaging past customers or subscribers who opted in to calls.
- Practice: an AI role play buyer that lets new callers rehearse openers and objections without burning real leads.
The last use is underrated. Role play with an AI buyer costs nothing in reputation, and new callers can repeat the hard moments until they sound natural. It trains the human who then makes the real cold call.
For the wider question of what an AI should do alone and what needs a person, read my guide on AI SDR responsibilities and handoffs.
How to Add AI to Your Cold Calls: A 7-Point Check
- Decide the model first: AI voice for consented calls only, human voice for cold outreach.
- Map every number type in your list. Treat any mobile as covered by the consent rules.
- Scrub against the National Do Not Call Registry and your internal do-not-call list.
- Tell the prospect who you are and why you are calling in the first seconds.
- Check recording consent rules for the states you call before you transcribe.
- Keep a human check on the templates the AI uses for buyer emails, and spot-check what it sends.
- Measure meetings held and accepted by sales. Dials and talk time do not show pipeline.
Run the check on your current list before you buy any tool. Most teams find the data problem first: wrong numbers, old titles and no record of consent. Fix that and both models work better.
The list is where it starts. See how the providers compared in our test of lead generation database companies.
What Should You Measure in AI Cold Calling?
Measure meetings held and accepted by sales, whatever model you run. Dials and minutes show effort. Only held meetings show pipeline. Track the chain in order, so you can see where it breaks.
| Metric | Human caller with AI | AI voice agent |
|---|---|---|
| Connect rate | Calls answered by the right person | Calls answered by the right person |
| Conversation to meeting | Meetings booked per real conversation | Meetings booked per qualified call |
| Show rate | Meetings held out of booked | Meetings held out of booked |
| Accepted by sales | Meetings the closer marks as a real opportunity | Same, plus how many a human had to rescue |
| Admin time | Minutes after each call before the next dial | Not applicable |
| Risk signals | Do Not Call requests | Opt-outs, complaints, calls without recorded consent |
Our own reference point is simple: 3 meetings booked in one hour of calls, with the admin after each call handled by AI. Use your first 100 conversations as a baseline, then change one thing at a time.
What Are the Best AI Cold Calling Tools?
The best AI cold calling tool depends on the model you chose. For AI voice agents, the search results are led by builders such as Retell AI and Synthflow. For human callers, phone systems such as RingCentral and Dialpad now add AI transcription and summaries. RingCentral lists its AI Conversation Expert from $60 per user per month. The CRM and email step can run on the AI assistant your team already uses.
Pick the workflow first. Then pick the tool. Write down the five steps from the workflow section, mark which ones a person must own, and buy only for the steps left. Most teams need fewer tools than the demos suggest.
Want callers who already work this way? See our B2B appointment setting agency, where people make the calls and AI runs the operations around them.
Frequently asked questions
What is AI cold calling?
AI cold calling is the use of artificial intelligence in outbound sales calls. It can mean an AI voice agent that dials and talks. It can also mean a human caller, with AI doing research, call summaries, follow-up emails and CRM updates. The two models carry very different legal and reputational risk, so pick one on purpose.
Is AI cold calling legal in the US?
It depends on consent. Since the FCC ruling of February 8, 2024, AI-generated voices count as artificial under the TCPA. Telemarketing calls with an artificial voice to mobile numbers need prior express written consent. A human caller using AI tools in the background is not an artificial voice, though Do Not Call and state telemarketing laws may still apply.
How much does AI cold calling cost?
Voice agent platforms price by the minute. Retell AI publishes about $0.07 to $0.31 per minute, including voice infrastructure, text-to-speech, the language model and telephony. That is roughly $350 to $1,550 for 1,000 five-minute calls. Setup, testing and compliance work come on top, and consent failures can cost $500 per violation, up to $1,500 when willful.
Can AI replace cold callers?
Not for first contact with B2B buyers today. The law restricts AI voices on cold calls to mobiles, and buyers who never opted in react badly to a synthetic voice. AI can replace most of the admin around a call. Our human caller, with AI handling the rest, has booked 3 meetings in one hour of calls.
Sources
Federal Communications Commission, FCC Makes AI-Generated Voices in Robocalls Illegal: Declaratory Ruling announced February 8, 2024, AI-generated voices are artificial under the TCPA, checked September 25, 2026.
US Code, 47 U.S.C. 227: prior express consent for artificial voice calls to mobile numbers, $500 per violation, up to three times when willful, checked September 25, 2026.
Code of Federal Regulations, 47 CFR 64.1200: prior express written consent for telemarketing with an artificial voice, checked September 25, 2026.
Retell AI, Pricing: voice agents from about $0.07 to $0.31 per minute, voice infrastructure $0.055 per minute, telephony from $0.015 per minute, read September 25, 2026.
Salesforce, What is AI cold calling: AI that supports sellers on their calls, December 18, 2025.
RingCentral, AI cold calling guide: AI Conversation Expert from $60 per user per month, updated July 21, 2026, read September 25, 2026.
Federico Donatone, LinkedIn post of August 26, 2026: my assistant books 3 meetings in 1 hour while AI runs summaries, follow-up, recaps and CRM updates.
Disclosure: Growth Cab sells appointment setting with human callers, so read this guide knowing we are one of the options. The 600-lead data test was paid for by Prospeo, a data provider I partner with. Tool prices come from each vendor's own page. This article is general information and not legal advice.
