AI cold calling and appointment setting: what works, what is restricted, and where reps step in

AI cold calling means a voice agent phones people who have not asked to hear from you and tries to book a meeting. In the US and UK, the legal limits on that matter more than the technology. Calls made with an AI voice are treated as automated, which generally means you need specific prior consent. The designs that hold up call warm leads who agreed to be contacted. They honor an opt-out instantly and transfer live to a rep.

This is general information, not legal advice. Rules differ by country and state, and they change. Have counsel review your plan before you dial anyone.

#Why the label misleads

"Cold calling" implies strangers with no relationship to you. That is the highest-risk version of outbound, and it is the one where AI voice is most restricted. The useful work in appointment setting is rarely cold. It is a form fill from last night, a trial signup that stalled, or a lead that asked for a callback and was never called. A voice agent that reaches those people within minutes is a different and much safer product than one dialing a purchased list.

Our post on the AI SDR covers the wider outbound picture, including email and LinkedIn steps that do not carry the same call-consent exposure.

#United States: the TCPA

The Telephone Consumer Protection Act (TCPA) restricts calls made with an "artificial or prerecorded voice." In February 2024 the FCC adopted a declaratory ruling (FCC 24-17) stating that this phrase covers current AI technologies that generate human voices. Calls using them therefore need the prior express consent of the called party, absent an emergency purpose or an exemption. A footnote adds that when the call introduces an advertisement or constitutes telemarketing, the FCC's rules require prior express written consent.

The ruling also notes that the caller must identify the entity responsible for the call and, for telemarketing, must offer specified opt-out methods. It states plainly that there is no carve-out for technology that purports to be the equivalent of a live agent.

In practice that means records and exits.

Your consent record is the asset. For each number you call you want to be able to show who consented, when, to what wording, and to which kind of contact. A "contact us" form checkbox that says nothing about automated or AI calls is a weak record.

Opt-out must work during the call and afterward. The FCC rules require an opt-out mechanism on covered artificial or prerecorded telemarketing calls, and a number that opts out goes onto your do-not-call list. Consent revocation by any reasonable method must be honored within a short period set out in the rules.

The National Do Not Call Registry is a separate rule. It restricts telephone solicitations to registered residential subscribers, and it applies to live calls as well.

Some states have their own telemarketing laws on top of the federal rules. Calling hours, registration and consent standards vary, so a nationwide campaign needs checking state by state.

#United Kingdom: PECR and the TPS

In the UK, the Privacy and Electronic Communications Regulations (PECR) govern direct marketing calls. The ICO's guidance sets two regimes.

For live marketing calls, you must not call numbers on the Telephone Preference Service (TPS) or Corporate TPS without specific consent, you must screen your lists against them and your own suppression list, and you must say who is calling and display your number.

For automated marketing calls, you may call only with specific consent that covers automated calls. The ICO says general marketing consent, or consent for live calls, is not enough. You must include your name and a contact address or freephone number, and stop if consent is withdrawn.

The ICO has addressed the AI-voice question in a published reply to an organization. The software in question used a synthesized call-handler voice to open outbound marketing calls. The ICO said the system fell within PECR's automated calling definition and that, because the calls opened with a recording, the organization needed specific consent. The ICO has also said it is investigating a rise in automated and AI-enabled marketing calls. We have not found an ICO statement addressing a fully live, real-time conversational agent specifically, so the careful reading is to treat an AI voice marketing call as an automated call.

#Official sources

We read each of these while writing this section. Check them for changes before you rely on them.

• FCC, Declaratory Ruling FCC 24-17, "Implications of Artificial Intelligence Technologies on Protecting Consumers from Unwanted Robocalls and Robotexts" (CG Docket No. 23-362, adopted February 2, 2024) (opens in a new tab)
• 47 CFR 64.1200 (TCPA delivery restrictions, do-not-call, opt-out) (opens in a new tab)
• ICO, telephone marketing guidance under PECR (opens in a new tab)
• ICO, "Previously asked questions" (AI-synthesised voice and the automated calling definition) (opens in a new tab)
• ICO news, automated and AI-enabled calls (October 2026) (opens in a new tab)

#What the rules mean for design

#A compliant design for AI appointment setting

1. Capture consent at the source. The form, chat or booking page states in plain words that the person may receive a call from an automated AI voice, and records the timestamp, wording version and number.
2. Gate every dial on the consent record. The agent cannot place a call to a number with no matching entry. Suppression and do-not-call lists are checked immediately before each call.
3. Open with identity and disclosure. The agent says who it is calling for, that it is an AI assistant, and why the call is happening. The agent should never pose as a named human or clone a real person's voice without clear disclosure.
4. Offer a clean exit at any moment. "Stop," "don't call me again," "remove me" or any reasonable equivalent ends the call, writes the suppression record at once and syncs it to your CRM and dialer.
5. Keep the goal narrow. The agent confirms interest, asks two or three qualifying questions and offers meeting slots from the rep's real calendar. It does not negotiate price or promise outcomes.
6. Transfer live. If the person wants to talk to someone now, the agent warm-transfers to an available rep with a summary. If no rep is free, it offers a specific callback time and creates the task.
7. Respect timing. Use the local calling windows that apply in each region and honor the person's stated preferred time.
8. Log everything. Store the recording or transcript, the consent record, the call outcome and every CRM write, so you can answer a complaint.

Key point: In outbound, consent records and opt-out handling are the product. The voice is a detail. Build the consent gate and the suppression sync before you tune a single prompt.

#Where reps step in

Reps take over for pricing, for objections beyond a basic answer, and for anyone who sounds annoyed or asks for a person. Anything the agent is unsure about goes to them too. Agent-booked meetings need a good brief: the lead's source, what they said on the call, their answers to qualifying questions and the consent record link.

For example, a software company with a demo-request form could have the agent call each new request within a few minutes. The agent confirms the person still wants a demo and books a slot. A calendar invite follows. A rep who gets the booking sees the transcript summary and the product the lead asked about. Leads who say they are not interested are closed out and suppressed. Our guide to AI lead generation shows how this fits with the earlier funnel stages.

#What to measure

Count meetings held, because booked meetings flatter the agent. Track no-show rate, opt-out rate per hundred calls, complaint count, and the share of calls where the lead asked for a person. A rising opt-out or complaint rate means stop. Don't optimize around it. Listen to a sample of calls every week.

#When not to do this

Skip outbound AI calling if you cannot show consent for the numbers you plan to dial. Skip it too if your sales cycle hinges on a long, relationship-led first call, or your market is small enough that one angry prospect matters. A well-written email sequence and a prompt human follow-up may beat a voice agent here. For inbound reception, which carries far less risk, see our post on the AI receptionist. If your team is in the UK, our UK page covers how we work there.

For the build side, see voice agents. If you want us to review a proposed outbound flow against your consent records, book a strategy call.

#Frequently asked questions

#Is AI cold calling legal in the US?

It depends on consent and the type of call. The FCC has ruled that calls using AI-generated voices fall under the TCPA's "artificial or prerecorded voice" rules, so they generally need prior express consent, and telemarketing calls need prior express written consent. Calling people who never consented is the high-risk case. Ask a lawyer to review your specific campaign.

#Is AI cold calling legal in the UK?

Automated marketing calls need specific consent that covers automated calls, and the ICO has said general marketing consent is not enough. The ICO has treated software that plays a synthesized voice in marketing calls as an automated calling system. Live marketing calls also must be screened against the TPS. Check with a specialist before dialing.

#What is an AI appointment setter?

It is a voice agent that calls or answers leads, confirms interest, asks a few qualifying questions and books a meeting on a rep's calendar. The better builds work from leads who opted in, write results to the CRM and transfer live when the person wants to speak to someone.

#Can an AI voice agent pretend to be a human on a call?

It should not. The FCC ruling requires callers to identify the responsible entity, and good practice is to tell the person it is an AI at the start. Impersonating a real person or cloning a voice without disclosure carries legal and reputational risk.

#Should AI make cold calls to a purchased list?

We don't recommend it and don't build for it. Without a consent record for each number, you risk breaching the TCPA in the US and PECR in the UK, and the complaints that follow can cost far more than the meetings are worth.