Short answer: AI outbound calling means two different things. AI voice agents dial the number and talk to the prospect themselves. AI-assisted dialing keeps a human rep on the call and uses AI to dial several lines at once, skip voicemail, write the notes and coach the rep. In B2B, voice agents fit short, structured calls to people who asked to hear from you: a form fill, a webinar sign-up, a meeting reminder, a customer renewal. Cold calls to new accounts still work best with a person talking and AI doing everything around the call. Before any AI voice dials a number, know that in February 2024 the FCC ruled that AI-generated voices count as “artificial” voices under the TCPA, and check consent with counsel.
A note on who ranks for this term. When we read the Google results on 3 November 2026, the top pages were a dialer vendor’s blog, a no-code voice agent builder, a Reddit thread about booking haircuts by phone, a healthcare billing company and a YouTube tutorial. None of them separated voice agents from AI dialers, and none gave a B2B team a way to test either. We do not sell calling software. We run outbound programs for B2B companies, so this guide is about where each kind of AI belongs in that work.
If you want the wider picture first, our guide to AI sales agents covers agents after the first meeting, and our AI BDR guide covers AI-run email and LinkedIn prospecting. This page is only about the phone.
What are the two kinds of AI outbound calling?
| AI voice agent | AI-assisted human dialing | |
|---|---|---|
| Who talks to the prospect | Software, with a synthetic voice | Your rep |
| What the AI does | Dials, listens, answers, books or transfers, logs the call | Picks who to call, dials several lines in parallel, skips voicemail and phone trees, transcribes, summarizes, coaches |
| Best at | Short calls with a clear goal: confirm, qualify, book, remind | Cold calls, discovery, objections, senior buyers |
| Main gain | Calls happen at any hour without a person | More live conversations per rep hour |
| Main risk | Consent, a bad answer said out loud, buyer trust | Spam-flagged numbers, reps rushing through connects |
| Example vendors (from their own sites, 3 November 2026) | Retell AI, Bland, 11x (Julian) | Orum, Nooks, Trellus |
Many teams end up using both: a dialer for the reps who prospect, and a voice agent for a few narrow jobs where speed matters more than conversation.
How does an AI voice agent handle an outbound call?
A voice agent chains three pieces of software in near real time: speech recognition turns what the prospect says into text, a language model decides what to say next, and a text-to-speech voice says it. Around that sit the telephony (the phone number and the call itself) and the integrations that read the contact from your CRM before the call and write the outcome after it.
A typical B2B call runs like this:
- Trigger. A contact enters a list: a demo request, a webinar sign-up, a meeting tomorrow, a renewal in 60 days.
- Context. The agent pulls the name, company, what they asked for and any open deal from the CRM.
- Call. It says who it is and which company it calls for, states the reason, and asks two or three questions you wrote.
- Outcome. It books a meeting, confirms one, records an answer, or transfers to a person. Bland, for example, prices “transfer time” separately from talk time on its pricing page, which tells you transfers to humans are a normal part of the flow.
- Log. The transcript, the answers and the next step go back to the CRM, and the call is flagged if the prospect asked for a person or to stop calls.
Vendors aim these agents at narrow jobs. 11x, for instance, describes its voice agents on its home page as built to qualify and convert inbound demand and to call existing customers about upgrades, routing to human reps to close, and lists “Consented Outbound Calling” as a platform feature. That framing, consented calls and a hand-off to a person, matches where we see voice agents work in B2B.
How does AI-assisted dialing work?
The rep stays on the phone. The AI handles the waiting and the typing.
- Parallel dialing. The dialer calls several numbers at once and connects the rep to whoever answers. Orum’s FAQ says that if two people pick up, it detects who answered first “down to the millisecond,” connects the rep to that person and cancels the other call (orum.com).
- Screening. The dialer skips voicemail, phone trees and dead numbers so the rep only hears live people. Orum describes its Operator agent as handling screeners and phone trees before connecting a live person.
- Number health. Dialers rotate caller IDs to limit spam flags. Orum’s pricing page lists up to 10 caller IDs per month in its base plan.
- Notes and summaries. Calls are recorded and summarized into the CRM. Orum lists AI Call Summaries and AI Objection Detection in its base plan.
- Coaching. AI scores calls and runs practice sessions. Orum sells AI Roleplay Agents, AI Scorecards and a Coaching Portal as an AI Coaching add-on, and Nooks lists AI-powered coaching among its products (nooks.ai).
Vendors make big claims about the gain. Nooks, for example, says its AI Dialer helps reps “connect with 3x more prospects.” We have not tested that number, and your connect rate depends on your list more than your dialer. Measure it on your own data in a pilot.
Where does each kind fit in a B2B outbound motion?
The deciding questions are whether the person expects your call and how much judgment the call needs.
| Call type | Prospect expects the call? | Judgment needed | Our default |
|---|---|---|---|
| Cold call to a new account | No | High | Human rep on an AI-assisted dialer |
| Follow-up on a demo request or pricing form | Yes, they asked | Low to medium | AI voice agent within minutes, transfer to a rep for anything beyond booking |
| Webinar or event follow-up | Usually, if they registered | Low | AI voice agent with two qualifying questions, or a rep if the account is a target |
| Meeting confirmation and no-show rebooking | Yes | Low | AI voice agent |
| Checking a contact is still at the company | Mixed | Low | AI voice agent on contacts with consent, otherwise a rep with a dialer |
| Expansion or renewal call to a customer | Yes | Medium | AI voice agent opens and qualifies, account manager closes |
| Discovery, multi-threading, negotiation | Yes | High | Human, with AI notes and coaching |
Two patterns explain the table. First, the inbound follow-up rows are speed problems. A study published in Harvard Business Review audited 2,241 US companies with a test lead and found only 37% responded within an hour; firms that tried to contact a lead within an hour were nearly seven times as likely to qualify it as firms that waited an hour longer. A voice agent that calls a demo request at 9 pm closes that gap. Second, the cold call row is a judgment problem. A new buyer decides in the first seconds whether you are worth their time, and a misheard answer or an invented fact said out loud cannot be edited the way an email draft can.
What does the FCC ruling mean for AI voice calls?
This section states one fact and gives no legal advice.
On 2 February 2024 the Federal Communications Commission adopted a Declaratory Ruling (FCC 24-17, CG Docket No. 23-362, released 8 February 2024). In it, the FCC confirmed that the TCPA’s restrictions on the use of “artificial or prerecorded voice” encompass current AI technologies that generate human voices. As a result, the ruling says, such calls require the prior express consent of the called party, absent an emergency purpose or an exemption.
What that means for your list depends on facts we cannot judge from here: what consent you hold and how you collected it, whether the numbers are business lines or mobiles, which states and countries you call, how you record calls, and how opt-outs are handled. Check with counsel before an AI voice agent dials anyone, and get the answer in writing for each list you plan to use. The same goes for AI-assisted dialers, which have their own rules on automated dialing.
What we do on every AI voice program, as operating practice rather than legal advice: the agent says at the start that it is an automated assistant and names the company, it offers a person on request, and it honors a stop request on that call and in the CRM.
What does AI outbound calling cost?
Voice agent platforms publish usage prices. AI dialers mostly do not. Prices below were read on each vendor’s own pricing page on 3 November 2026, in US dollars.
| Vendor | Kind | Published price | Notes |
|---|---|---|---|
| Retell AI | Voice agent platform | $0.07 to $0.31 per minute for voice agents, pay as you go | $10 in free credits; 20 concurrent calls free, then $8 per concurrent call per month; phone numbers $2 a month; branded calls +$0.10 per outbound call |
| Bland Start | Voice agent platform | $0.14 per minute of talk time, $0.05 per minute of transfer time, $0 platform fee | Caps of 100 calls a day and 10 concurrent calls |
| Bland Build | Voice agent platform | $299 a month, then $0.12 per minute talk, $0.04 per minute transfer | Caps of 2,000 calls a day and 50 concurrent calls |
| Orum | AI-assisted dialer | Not published, “Request pricing” | Minimum 9 seats; AI Coaching is an add-on |
| Nooks | AI-assisted dialer and coaching | Not published, custom quote | Sells dialer, sequencing, coaching and a data package |
| 11x | Voice agent (Julian) | Not published, demo only | Voice agent aimed at inbound, expansion and consented outbound |
Per-minute prices are only part of the bill. The model behind the agent, the voice, the phone numbers, the CRM connection and, above all, the person who reviews calls every week all cost money. Retell’s page lists separate per-minute rates for each voice and language model, so the same agent can cost several times more depending on the setup.
Illustrative cost of a small pilot (not a quote; the numbers are made up to show the math): 400 calls a month to demo requests and webinar sign-ups, of which 250 connect, averaging 3 minutes each, is 750 talk minutes.
| Platform and plan | Calculation | Usage cost per month |
|---|---|---|
| Bland Start | 750 x $0.14 | $105 (fits under the 100-calls-a-day cap) |
| Bland Build | $299 + 750 x $0.12 | $389 |
| Retell AI, low end | 750 x $0.07 | $52.50, plus $2 per phone number |
| Retell AI, high end | 750 x $0.31 | $232.50, plus $2 per phone number |
At this volume, software is cheap next to the setup and review time. Budget for someone to write the call flow, connect the CRM and listen to a sample of calls every week. For a comparison with the cost of a human rep, our SDR cost calculator models salary, tools and ramp time.
How do you pilot AI outbound calling?
Run one narrow pilot for four weeks before you commit to a platform or a dialer contract.
- Pick one call type from the table above. Demo-request follow-up is the usual first choice: the contact asked to hear from you, the goal is a booked meeting, and you can compare against how fast your team replies today.
- Settle consent first. Write down which contacts the pilot will call and what consent you hold for each. Have counsel sign off on the list and the opening line.
- Write a short call flow. The opening (who is calling and why, that it is automated), two or three questions, the booking step, and the hand-off rules: pricing, contract terms, complaints and anything off-script go to a person.
- Connect the CRM before the first call. The agent should read the contact and write the outcome, transcript and opt-outs. A pilot with results in a spreadsheet tells you nothing about running it at scale.
- Listen to calls every day in week one. Read or hear every transcript for the first 50 calls, then a sample of 20 a week. Fix the script and the knowledge the agent draws on each time it says something wrong.
- Run a control. Keep a share of the same contacts on your current process, or compare with the previous month, so you can tell whether the agent changed anything.
- Decide at week four using the metrics below, and write down the decision and the reason.
For a dialer pilot, the steps are the same with one change: split reps, not contacts. Put two reps on the AI dialer and two on the current setup for the same list type, and compare conversations and meetings per hour of calling.
Which metrics show that AI outbound calling works?
| Metric | How to measure | Red flag |
|---|---|---|
| Speed to first call | Minutes from form submit to first dial, from CRM timestamps | Slower than your team’s best hour |
| Connect rate | Live conversations / dials | Falling week on week (numbers getting flagged) |
| Booking rate | Meetings booked / live conversations | Lower than your reps on the same list type |
| Meeting held rate | Meetings held / meetings booked | AI-booked meetings no-show more than rep-booked ones |
| Transfer rate | Calls passed to a person / live conversations | Near zero (agent answering what it should not) or very high (script too narrow) |
| Opt-out and complaint rate | Stop requests and complaints / live conversations | Any complaint about not being told it was AI |
| QA error rate | Reviewed calls with a wrong fact or wrong action / calls reviewed | Above a few percent after week two |
| Cost per held meeting | Software + review hours / meetings held | Higher than the rep cost per held meeting |
The two that matter most are meeting held rate and the QA error rate. A voice agent that books many meetings nobody attends, or that books them by saying things your sellers then have to walk back, costs more than it saves.
How does this differ from AI SDRs and AI BDRs?
The names overlap, so here is how we use them. An AI SDR follows up on inbound leads, usually by email, chat and calendar link; a voice agent calling demo requests is an AI SDR that uses the phone. An AI BDR prospects accounts that never raised their hand, mostly by email and LinkedIn, because cold AI voice calls carry the consent and trust problems above. AI sales agents is the wider family, including agents that brief calls, draft recaps and update the CRM after a meeting.
Phone is one channel in an outbound sequence, next to email and LinkedIn. Our guide to what outbound sales is covers how the channels fit together.
Where Growigami fits
We do not resell calling software. Our ABM and outbound service runs outbound for B2B companies with AI doing the research and the drafts and our team reviewing before anything reaches a buyer. Where phone belongs in the plan, we set up the dialer or the voice agent inside your CRM, run the four-week pilot above, and report the metrics in the table each week. If your bigger gap is slow replies to inbound leads, our inbound lead follow-up service covers that first.
Frequently asked
What is AI outbound calling?
Using AI to place or support outgoing sales calls. It comes in two kinds. An AI voice agent dials the number and holds the conversation itself, then logs the result or transfers the call to a person. An AI-assisted dialer keeps a human rep on the call and uses AI to dial several numbers at once, skip voicemail and phone trees, summarize calls and coach the rep.
Can AI make cold calls?
Technically, yes: voice agent platforms can dial lists and hold a scripted, adaptive conversation. Whether it is allowed is a separate question. In February 2024 the FCC ruled that AI-generated voices count as artificial voices under the TCPA, which brings consent requirements with them. For B2B teams, we use AI voice agents on contacts who asked to be called and keep cold calls with human reps on AI-assisted dialers. Check the consent rules with counsel before any AI voice dials a number.
Is AI outbound calling legal?
We cannot give legal advice. The fact to know is that the FCC's Declaratory Ruling FCC 24-17, adopted on 2 February 2024, confirmed that the TCPA's rules on artificial or prerecorded voices cover current AI technologies that generate human voices, so those calls need the prior express consent of the called party unless an exemption applies. What counts as consent, and how state and international rules apply to your list, is a question for counsel.
How much does AI outbound calling cost?
Voice agent platforms publish per-minute prices. On 3 November 2026, Retell AI listed voice agents at $0.07 to $0.31 a minute on pay as you go, and Bland listed $0.14 a minute of talk time on its free Start plan or $0.12 a minute on its $299-a-month Build plan. AI-assisted dialers such as Orum and Nooks did not publish prices; both ask you to request a quote, and Orum lists a 9-seat minimum.
What is the difference between an AI dialer and an AI voice agent?
Who talks. With an AI dialer, a rep talks and the software decides who to call, dials several lines in parallel, connects the rep only when a person answers and writes the notes. With an AI voice agent, software talks to the prospect. Dialers raise the number of live conversations per rep hour. Voice agents remove the rep from the call, which suits short, structured calls better than discovery.
Will prospects know they are talking to AI?
They should. We tell people at the start of every AI voice call that they are speaking with an automated assistant for a named company, and we offer a person on request. Disclosure rules vary by place, so confirm the exact requirement with counsel. Beyond the rules, a B2B buyer who finds out later that a 'rep' was a bot rarely books a second call.
