AI Cold Calling: What It Is, Legal Rules and a Checklist
AI on the Cold Call
AI cold calling explained: the three models, the TCPA rules that apply, where AI helps and where reps should stay, plus a pre-launch checklist.

On this page
- What is AI cold calling?
- Is AI cold calling legal?
- How do you collect and record consent for AI calls?
- Where does AI help in cold calling, and where should reps stay in charge?
- Example AI cold calling scripts
- How does Phonify handle AI cold calling?
- How do you set up an AI cold calling campaign?
- AI cold calling pre-launch checklist
- The honest limits of AI cold calling
Every sales team has the same problem: most outbound calls reach voicemail or a polite "not interested", and reps burn hours finding the few people worth talking to. AI cold calling aims at that waste. Used well, it screens the list so people talk to prospects who actually want a conversation. Used badly, it breaks consent rules and annoys the people you want as customers.
This guide explains the three models of AI cold calling, what the law says, where AI helps and where it should stay out, and gives you a checklist to run before your first AI campaign.
What is AI cold calling?
AI cold calling is the use of AI in outbound prospecting calls to people who have not asked to hear from you. It covers anything from tools that help a rep prepare and log calls to AI voice agents that dial a list, hold the first conversation and hand interested prospects to a person.
The term is used loosely, so it helps to separate three models:
| Model | What the AI does | Who talks to the prospect | Main risk |
|---|---|---|---|
| AI assists the rep | Research, call scripts, transcription, summaries, next-step suggestions | A human rep, always | Low; mainly data privacy and call recording consent |
| AI makes the first call | Introduces itself, asks two or three qualifying questions, then transfers interested prospects | AI first, then a rep | Consent, disclosure and a bad first impression if it is clumsy |
| AI runs the whole call | Handles the conversation start to finish, including booking | The AI | Highest; objections, trust and legal exposure |
Most teams should start with the first or second model. The third is only sensible for narrow, low-stakes calls, such as confirming appointments with existing customers.
Is AI cold calling legal?
It can be legal, but calls that use an AI voice carry extra obligations. In 2024 the FCC confirmed that AI-generated voices are "artificial" voices under the TCPA, which generally means you need the called person's prior express consent before placing those calls to mobile phones or residential lines. Some states add their own disclosure rules. The full text is the FCC's Declaratory Ruling FCC 24-17, adopted on February 2, 2024.
The underlying regulation, 47 CFR 64.1200, also sets what an artificial-voice message must say. Paragraph (b) requires it to state clearly, at the beginning, the identity of the business responsible for the call, and to give a phone number for that business during or after the message. Telemarketing messages to residential lines and mobile numbers must also offer an automated opt-out, by voice or key press, during the call.
Whatever the AI does, the basics of lawful outbound calling still apply:
- Check the National Do Not Call Registry and your own internal do-not-call list before dialing.
- Respect calling hours, which in the US means no telemarketing calls to residential numbers before 8 am or after 9 pm in the called party's local time.
- Honor opt-outs immediately, across every channel.
- Disclose the AI at the start of the call, and tell people when the call is recorded.
This is general guidance, not legal advice. Read the FCC's page on telemarketing and robocalls and check your plan with counsel. Our TCPA calling hours and DNC checklist turns the day-to-day checks into a list.
How do you collect and record consent for AI calls?
Collect consent in writing before the first AI call, in a form that names your business and says the person agrees to receive marketing calls that may use an artificial or AI-generated voice, then store that record where the dialer can check it. For telemarketing calls that use an artificial voice, to mobile numbers or residential lines, the rule asks for prior express written consent.
Under 47 CFR 64.1200, a written agreement for this kind of call must clearly tell the person that signing authorizes the seller to deliver such calls, and that they do not have to agree as a condition of buying anything. An electronic signature, such as a checked box on a web form, can count where federal or state law recognizes it.
- Write the consent language. Name your business, the kind of calls (marketing calls that may use an AI or artificial voice and automated dialing), the number they will reach, and that agreeing is not a condition of purchase. Have counsel review it.
- Capture it on an unchecked box. The person must take an action; a pre-ticked box or a line buried in general terms will not hold up.
- Record the evidence. Store the date and time, the page or form, the exact wording shown, the phone number entered and the business it names.
- Check before every dial. The campaign should skip any number without a matching, unrevoked record.
- Honor revocation everywhere. A "stop calling" on an AI call, a STOP text or an email request should reach every campaign and channel. The FCC sets a maximum time for processing revocations; check the current text of 47 CFR 64.1200(a)(10) and aim to act faster.
An example of consent wording, as an illustration to adapt with counsel: "By checking this box, I agree that Northwind may call me at the number above with marketing calls, including calls that use automated dialing or an AI-generated voice. Agreeing is not a condition of purchase. I can opt out at any time."
Where does AI help in cold calling, and where should reps stay in charge?
AI helps most with repetitive, low-judgement work: screening a list, asking a few fixed questions, logging outcomes and summarizing calls. Reps should own anything that needs judgement or trust: objections, pricing, negotiation and closing.
| Task | AI | Rep |
|---|---|---|
| Working through a large call list | Strong fit | Slow and tiring at scale |
| Asking "are you the right person?" and two qualifying questions | Good fit | Fine, but a poor use of time |
| Voicemail and no-answer handling | Good fit | Tedious |
| Call notes and summaries | Strong fit | Often skipped |
| Handling objections | Weak fit | Strong fit |
| Pricing, negotiation, closing | Not recommended | Owns it |
| Upset or confused prospects | Hand off immediately | Takes over |
The pattern is simple: AI does the sorting, people do the selling. The link between the two is the transfer. A prospect who says "yes, tell me more" should reach a rep in seconds, with the reason for the call already known. We cover the policy side in our guide to AI to human handoff.
Example AI cold calling scripts
An AI caller's script should be short, open with a disclosure and the business name, and make it easy to reach a person or opt out. The lines below are illustrations with an assumed company; write your own with your counsel and test every branch.
Opening with disclosure
"Hi, this is an AI assistant calling for Northwind, and this call may be recorded. You asked about our phone system on our website last week. Do you have a minute for two quick questions, or would you prefer we call back later?"
Handoff to a person
"Thanks, that helps. I can connect you with Sam on our team now, who can answer pricing questions. Is that OK?" If nobody picks up: "Sam is on another call. Can we call you back today, or would tomorrow morning suit you better?"
Opt-out and "is this a robot?"
"Understood. I will make sure we do not call you again. Our number is [business number] if you ever want to reach us. Goodbye." And if asked: "Yes, I am an AI assistant. Would you like to speak with a person instead?"
Every opt-out line should end the call and record the request at once. If the call reaches voicemail instead, prerecorded or AI messages raise the same consent questions; our guide to voicemail drop covers how those messages work.
How does Phonify handle AI cold calling?
Phonify combines a power dialer with outbound AI campaigns on the same account: an AI voice agent can call a list, open with an AI disclosure, and warm-transfer interested prospects to a rep or ring group. Consent, Do Not Call and calling-hours checks sit in the same platform. Here is what that covers, based on the live product pages:
- Outbound AI campaigns. An AI agent calls a list and warm-transfers interested prospects to a user, ring group, number or queue.
- AI disclosure and recording notice. Outbound AI calls open with a disclosure, in your words or the default, and each agent keeps its own text.
- Compliance checks. Consent records, Do Not Call checks and calling-hours checks are included, though compliance remains your responsibility.
- Limits on data. You choose which lead fields an agent can use. Phone numbers, emails, SSNs, dates of birth and card or account numbers are always held back.
- After the call. AI-handled calls come with a transcript, summary, sentiment and a suggested outcome, plus QA scoring.
- Human calling on the same lists. Reps use the power dialer for the calls that need a person, and the AI receptionist uses the same agent technology for inbound calls.
Pricing is published. AI voice agents and outbound AI campaigns are on the Enterprise plan at $29.99 per user per month, and AI agent calls cost $0.06 per connected minute, paid from a prepaid balance. Unanswered calls cost nothing. See the plans and published rates for current numbers.
What Phonify does not claim: it does not replace your reps, and it does not make a campaign compliant by itself. You still need valid consent and a list you are allowed to call.
How do you set up an AI cold calling campaign?
Start with one narrow list and one clear goal, and review real calls before you scale. Follow these steps:
- Pick one goal. For example: find out whether the prospect is the right contact and interested in a demo.
- Choose a clean, permitted list. Confirm consent where it is required, and scrub against Do Not Call.
- Write the call flow. A short disclosure, one reason for calling, two or three qualifying questions, and a clear "transfer to a person" trigger.
- Set the handoff destination. A ring group or queue, not one person, and a fallback if nobody answers.
- Test. Run test calls that hit every branch, including a prospect asking for a human straight away.
- Pilot for two weeks. Listen to a sample of calls every day at first.
- Tighten and scale. Adjust the script from what you hear, then widen the list.
The AI voice agent setup checklist covers the configuration side, and our cold calling guide covers the human skills your reps still need.
AI cold calling pre-launch checklist
Run through this list before the first AI call goes out:
- Written consent or another lawful basis confirmed for every number on the list
- List scrubbed against the National Do Not Call Registry and your internal list
- Calling hours set by the called party's local time
- AI disclosure and recording notice written and tested
- List of topics the AI never handles, with an immediate handoff for each
- Handoff destination is a ring group or queue with a fallback
- Data the AI can see limited to what the call needs
- Opt-out phrases handled and recorded on the first request
- Daily call sampling scheduled for the first two weeks
- One owner named for the script and every change versioned
The honest limits of AI cold calling
AI cold calling does not fix a bad list, a weak offer or a missing consent record. It makes whatever you feed it faster, including mistakes. Prospects also differ in how they react to an AI caller, so measure your own results, such as transfers accepted and meetings booked, rather than relying on vendor claims, including ours.
If your team makes a small number of high-value calls, the better investment may be a good dialer and better research, not an AI caller. Our comparison of power dialer software helps you decide.
Frequently asked questions
What is AI cold calling?
AI cold calling is the use of artificial intelligence in outbound prospecting calls to people who have not asked to hear from you. It ranges from tools that help reps prepare and log calls, to AI voice agents that dial a list, introduce themselves, ask qualifying questions and pass interested prospects to a human.
Is AI cold calling legal?
It can be, but the rules are strict. In 2024 the FCC confirmed that AI-generated voices count as artificial voices under the TCPA, which generally requires prior express consent for those calls to mobile numbers. Do Not Call lists, calling hours and state laws also apply. This is not legal advice, so confirm your plan with counsel.
Does AI cold calling work better than human cold calling?
Not as a replacement. AI is good at high-volume, repetitive work such as screening a list, asking two or three qualifying questions and logging outcomes, so reps talk to fewer dead ends. Humans are better at objections, rapport and complex deals. Most teams get the best result by combining the two with a warm transfer.
Should an AI caller say it is an AI?
Yes. Opening with a plain disclosure that the caller is an AI assistant is the safest practice and is increasingly expected by regulators. It also protects trust: prospects who discover mid-call that they were talking to a machine tend to hang up and remember your brand for the wrong reason.
How much does AI cold calling software cost?
Pricing varies widely and many vendors only quote on request. Phonify publishes its rates: AI voice agents and outbound AI campaigns are on the Enterprise plan at $29.99 per user per month, and AI agent calls cost $0.06 per connected minute from a prepaid balance. Unanswered calls cost nothing.
Can AI replace SDRs?
Unlikely for most B2B teams. AI can take over screening, dialing and note-taking, which are the least rewarding parts of an SDR's day, but selling still depends on judgement, empathy and handling objections. A practical goal is more conversations per rep, not fewer reps.


