"AI dialer" is now on almost every calling vendor's homepage. The label covers at least four different products. One types up notes after a call. One decides who to call next. One dials faster than a human could. One puts a synthetic voice on the line that talks to your prospect. These carry very different legal risk, and the demo usually doesn't tell you which one you're buying.
This post is for sales leaders who have to make that call. It separates the parts of AI calling that already work from the parts that are mostly marketing. It covers where the TCPA and disclosure rules bite hardest. It also names the teams that should wait a year before putting a bot on the phone. We make Ready, which includes a dialer, and we'll be plain about what ours does and doesn't do.
The four things "AI dialer" can mean
Before you compare vendors, work out which of these you are actually looking at:
- Conversation intelligence. The call is recorded and transcribed. Software then produces a summary, action items or coaching notes. A human still dials and talks.
- AI-assisted list work. The software scores leads, suggests the best time to call, or reorders the queue. A human still dials and talks.
- Automated dialing marketed as AI. These are predictive or power dialers that place calls without a rep pressing a button. They often add answering-machine detection and call it "AI." We cover the mechanics in predictive dialer vs auto dialer.
- AI voice agents. A synthetic or cloned voice speaks to the person who answers, either from a script or generated live. No human is on the call.
Categories 1 and 2 change how your reps work. Categories 3 and 4 change how the law treats the call. That distinction should drive most of your evaluation.
What AI genuinely automates well today
Some of this is real and saves rep hours:
- Transcription and notes. Writing up a 6-minute call by hand takes a rep a few minutes. A transcript and summary dropped into the CRM takes none. Across 40 calls a day, that time adds up quickly.
- Disposition and field updates. Tagging a call "left voicemail" or "asked for callback Thursday" is a good fit for software, provided a rep can correct it.
- Coaching review. Searching 500 transcripts for "price objection" beats a manager listening to random recordings.
- Queue ordering. Putting fresh inbound leads ahead of 90-day-old ones is simple logic. You don't need a model for it, but it works.
Other claims are mostly hype, at least for most teams right now:
- "AI books meetings from cold calls." A voice agent can run a script. Handling an annoyed decision-maker, a gatekeeper or an unusual objection is a different job. And cold calling is exactly where the legal exposure is worst (see below).
- "Doubles your connect rate." Connect rates depend on list quality, caller ID reputation and timing far more than on any model. Ask for the vendor's methodology before you believe a multiplier.
- "Fully compliant out of the box." No dialer makes a call compliant. Compliance depends on who you call, what consent you hold and what the call says.
The compliance line: AI voices are "artificial voices"
This is the part that should slow you down.
The FCC already ruled on AI voices
In February 2024 the FCC issued a declaratory ruling. It confirmed that AI-generated voices, including voice clones, count as an "artificial or prerecorded voice" under the TCPA. In practice that means:
- Calls to mobile numbers using an AI voice need the called party's prior express consent. If the call is telemarketing, they need prior express written consent.
- Artificial-voice calls must identify the business at the start and give a way to opt out.
- A residential landline can't be called with an artificial voice without consent either, outside narrow exemptions.
Statutory damages run $500 per violating call, up to $1,500 if willful. A voice agent that works through 2,000 numbers in an afternoon can turn one bad list into a large claim in a single day. We cover the general framework in prior express written consent vs implied consent.
DNC still applies, and so do litigator lists
An AI voice doesn't exempt you from the National DNC Registry, state DNC lists or your own internal do-not-call list. Separately, some numbers belong to people who file TCPA suits professionally. Those three lists do different jobs, and confusing them is how scrubbed lists still get sued. See the three separate lists.
Disclosure: telling people it's a bot
Several states regulate undisclosed bots in sales contexts. California's bot disclosure law is the best-known example, and more AI-voice bills keep appearing. The rules vary and change quickly, so have counsel check the states you call into. The safe default is simple: say in the first sentence that the caller is an automated assistant.
Recording consent
Every AI voice agent records the call, because it can't function otherwise. In all-party-consent states, everyone on the call must be told before recording starts. That disclosure is one more sentence the bot has to say before it can begin selling.
Opt-out has to cross channels
If a prospect tells your voice agent "stop calling me," that request needs to reach your texting and email tools too. Most stacks don't sync opt-outs across channels by default. We wrote about why opt-out rarely crosses channels and how to fix it.
Risk by feature type
| Feature | What it changes | TCPA exposure | Fits |
|---|---|---|---|
| Transcripts and summaries | Rep admin time | Low (recording consent still applies) | Almost any team |
| Lead scoring / queue order | Who gets called first | Low | Teams with real lead volume |
| Automated or predictive dialing | Calls placed without a rep pressing dial | Moderate to high, depending on system and consent | Large teams with consented lists and counsel |
| AI voice agent | No human on the call | High: artificial-voice rules apply | Consented, inbound-origin leads only, with disclosure |
Who should wait
Hold off on AI voice agents and automated dialing if any of these describe you:
- You call cold or purchased lists. For an artificial voice, cold means no consent, which means each call is a potential violation. Cold outreach belongs with a human on the phone, after DNC and litigator scrubbing.
- You can't produce a consent record per contact. That means when the lead opted in, what wording they saw and on which page. If your forms or lead vendors don't capture this, fix that first.
- You're in a regulated vertical. In healthcare, lending, insurance and debt collection, the extra rules stack on top of the TCPA.
- Your team makes under roughly 100 dials a day. At that volume, a rep dialing manually plus automatic transcription gets most of the benefit with almost none of the artificial-voice exposure.
- You can't say which states you call into. Disclosure and recording rules depend on the recipient's state. You need to know where your contacts are.
The teams that are reasonable candidates look different. They work leads that came to them: form fills, inbound calls, existing customers. They have documented consent and a lawyer who has read the vendor's script.
A worked example: the low-risk version of "fast follow-up"
Much of the AI-dialer pitch is about speed to lead. You can get most of that speed without a bot speaking to anyone. Here is the math for a manual setup on Ready. These are example assumptions, not measured results.
Assumptions: 200 inbound web leads a month, all of whom checked a consent box. Each gets an instant 1-segment text ("Thanks for reaching out, we'll call you in a few minutes"). A rep then calls each lead from the dial pad. 60% answer, including voicemail pickups. The average answered call runs 4 started minutes.
- Texts: 200 × 1 segment × ($0.0075 + $0.0045) per segment.
- Calls: 120 answered × 4 minutes × $0.025/min = $28.80
- Unanswered calls: 80 × $0 = $0. Ready only bills answered time. A ring-out or busy line costs nothing, while a voicemail pickup counts as answered.
- Call cost per connect: $28.80 ÷ 120 = $0.24
Ready's all-in prices per segment include the carrier fee but not screening: Scrub lists and Number validation, the two switches in Settings under Compliance, each add $0.003 per segment when switched on, shown as screening on your bill, and both are off by default. Unless you turn the opt-out footer off, the first text to each contact, and one again from time to time, also gets "Reply STOP to opt out" and your business name if the message does not already have them, so a text close to the segment limit can bill one more segment.
Optionally, before calling, you could run a one-time standalone TCPA and DNC litigator scrub on those 200 numbers. At $0.005 per contact, that's $1.00.
To be clear about what this includes: the Ready Dialer is click-to-call from the pad in the dashboard or GoHighLevel. It includes consented call recording and transcripts, inbound calls that ring the open pad, and a Missed call card in the Inbox thread for each call you miss. It has no AI voice agent, no auto dialing and no voicemail drop. If you need a bot talking on your behalf, we aren't the tool, and that's fine. For a sales team working consented inbound leads, a human calling within five minutes after an instant text is a strong setup, and it carries little artificial-voice risk. Texting rates are on our pricing card, and the dialer is $0.025/min with no seat plans.
Questions to ask any AI dialer vendor
- Does the system place calls without a rep's action? If so, what is your position on ATDS and artificial-voice rules?
- How does the voice agent disclose that it's automated, and can we edit that line?
- Where is the recording disclosure played, and is it played before anything else?
- When someone says "stop" or "don't call me," where does that opt-out go, and does it reach our SMS tool?
- What consent evidence does your system store per contact, and what do we have to supply?
- Who carries liability if a call violates the TCPA? Read the indemnity clause, not the sales deck.
For the non-AI basics of picking a dialer, such as connect rates, caller ID and CRM sync, see our sales dialer guide.
The practical takeaway
Treat "AI dialer" as two separate purchases.
- AI that helps reps (transcripts, summaries, queue ordering) is low risk and saves time today.
- AI that replaces reps on the line is an artificial voice in the eyes of the FCC. It needs consent, disclosure, recording notice and opt-out handling, and it is a poor fit for cold lists.
If you're unsure, start with the low-risk version: instant text, a human call within minutes, recorded and transcribed with consent. Measure cost per connect for a quarter. If you want to try that setup, you can create a Ready account. Your first 20 one-to-one texts are free, and calls bill only when someone picks up.