Cloud Based Auto Dialer and Cloud Based Predictive Dialer Pricing

How much does a predictive dialer cost?

Between roughly $95 and $240 per agent per month if the vendor publishes a price, and unquotable if it does not. The spread is not quality. It is what the seat includes.

At the low end, BatchDialer publishes $95 per agent a month on annual billing for its Starter plan, which is $1,142 a year per agent, and $199 for its top tier. It bundles three to five dialing lines per agent and, unusually, includes DNC and Litigator Scrub on the entry plan. PhoneBurner sits at $140 to $183 per user a month on annual billing with unlimited calling, but it is a single line power dialer with no predictive mode at all, so it is priced against a different job.

At the enterprise end, Genesys Cloud CX publishes $75 for CX 1, $115 for CX 2, $155 for CX 3 and $240 for CX 4, per named user a month billed annually, and all four tiers list outbound campaigns as included. Five9 publishes $119 for Digital and $159 for Core, then adds the sentence that changes the arithmetic: prices above are per concurrent user and usage-based pricing may apply, with a minimum of 50 seats. Fifty Core seats is $7,950 a month before telephony, whether you dial for 5,000 minutes or 50,000.

Convoso publishes nothing, sells annually and offers no free trial, so it cannot be budgeted from its website at all.

Then there is the floor nobody puts on a pricing page. Genesys' own documentation recommends at least 15 agents in a predictive campaign. Fifteen CX 1 seats at $75 is $1,125 a month, or $13,500 a year, before a single minute of carrier traffic. That, not the $75, is the real entry ticket for Genesys predictive dialing. We break the same exercise down by dialing mode on our predictive dialer software page and against per-minute AI platforms on auto dialer pricing.

Cloud based dialer pricing: seats, concurrency or interacting hours

Three billing shapes exist in this category and vendors rarely explain the difference, which is a shame because it is worth more than any discount you will negotiate.

A named seat is a licence tied to one identifiable person. Simple to administer, and you pay for it whether that person is on the phone or on holiday. A concurrent seat is a licence for one simultaneous active user, so a 24-hour operation running three shifts of 20 reps can buy 20 concurrent licences rather than 60 named ones. That is why Five9's $159 looks cheaper than it reads: concurrent pricing usually costs more per licence but fewer licences, and Five9 also notes that its workforce engagement tooling is sold per named agent, so 50 concurrent seats spread across 120 staff still buys 120 WEM licences.

The third shape is the interesting one for outbound teams and almost nobody talks about it. Genesys' pricing page lists a licence type called Hourly Interacting and describes it as charging only for time spent handling interactions. For a campaign that runs four hours a day, five days a week, that is a fundamentally different bill from a named seat that sits idle the rest of the month. Genesys does not publish a rate for it, and every tier button on the page says Get custom pricing regardless of which licence type is selected, so you have to ask. Ask anyway.

On top of the licence there is nearly always telephony. Per-minute carrier cost, DID rental, recording storage and sometimes a separate charge for AI features. Genesys CX 4 includes 30 AI Experience tokens per named agent per month on top of a base allocation of 250 named tokens per organization, which is a consumption meter sitting inside a per-seat contract. Model the two lines separately or the forecast will be wrong. We ran the same seats-against-meters comparison across the enterprise platforms on outbound call center software.

Hosted predictive dialers: pros and cons

The honest case for a hosted or cloud predictive dialer is that it removes a job most sales teams should not be doing. No servers to size, no Asterisk to patch, no capacity planning before a campaign, no on-call rotation when a dialer stops dialing on a Tuesday morning. You get a browser, an admin console and a support contract, and you can add ten agents next month without buying hardware.

The cons are real and they cluster in three places. First, the meter never stops: a per-seat contract bills for reps who are ramping, on holiday or between campaigns, and an annual commitment locks that in. Second, you inherit the vendor's carrier relationships and their reputation, which cuts both ways: a good provider gets you A-level attestation and clean numbers, a bad one puts your calls behind somebody else's spam history. Third, you lose the last mile of control. If you need a pacing behaviour, a dialplan quirk or a data residency arrangement the vendor does not offer, the answer is no.

Self-hosting inverts all three. VICIdial is Open-Source AGPLv2 licensed, with no software licensing cost, per the project's own site, and it reports over 24,000 installations in over 100 countries, several with more than 300 agent seats. So the model demonstrably works at scale. What it costs you is infrastructure and expertise: servers, SIP trunks, DIDs, recording storage, and someone who understands both Asterisk and the compliance stack. A predictive dialer also needs meaningfully more concurrent channels than it has agents, because it places more calls than it has reps to answer them, so trunk sizing is a real design exercise rather than a checkbox.

The middle option, a third party that hosts VICIdial for you, is genuinely popular and worth a look if you want the software's flexibility without racking anything. Just be clear about who owns the carrier relationship in that arrangement, because that is what sets your attestation, and read the predictive dialer against power dialer comparison before you decide you need predictive at all.

How does a predictive dialer work?

A predictive dialer places calls before a rep is free, using a model of how long calls last and how often they connect, so that a live person is on the line at roughly the moment a rep becomes available. Genesys describes its own version as a patented stage-based algorithm that takes average handle time, after-call time, idle time, the stage setting on the agent script and historical contact rate as inputs.

That is why it needs agents. Genesys' documentation is direct about it: due to the difficulty of calculating average handle time, after-call time, idle time and contact rate for a small sample size, it recommends at least 15 agents in a campaign using the predictive dialing mode, and at least 10 to 15 for power mode. It then adds a caveat almost nobody accounts for, that an agent logged into two campaigns is not a whole agent in either. Its example: if 20 agents log in to both Campaign A and Campaign B, each campaign effectively has only 10 agents, which is under the recommended minimum.

The cost of guessing wrong is abandoned calls, and abandonment is regulated. Under the FTC's Telemarketing Sales Rule at 16 CFR 310.4(b)(1)(iv), a call is abandoned if a person answers and the telemarketer does not connect them to a sales representative within two seconds of their completed greeting, and the safe harbor requires abandoning no more than three percent of all calls answered by a person, measured over a single campaign if it runs less than 30 days, or separately over each successive 30-day period. That three percent is a budget, not a target, and it is why Abandon Rate is a configurable campaign property on power and predictive modes but not on preview or progressive.

So the practical decision tree is short. Under about ten dialing reps, a power or progressive dialer will give you most of the efficiency with structurally less abandonment risk, which is what our power dialer software page covers. Above fifteen, predictive starts to earn its keep. If your constraint is that you do not have fifteen reps and are not going to hire them, the honest answer is that predictive pacing is not your problem to solve, and running parallel dialer software or moving the conversation to an AI agent addresses it from the other direction.

Genesys cloud dialer: what the six modes actually do

Genesys Cloud documents six outbound dialing modes, and picking the wrong one is the most common self-inflicted cost in this category.

Preview shows the rep the record before dialing and the rep places the call, which Genesys lists as eliminating call abandonment at the cost of being the least efficient mode. Progressive dials one contact per available agent, and offers a precise mode that reserves the agent before the call is placed so they cannot be pulled onto an inbound call mid-dial. Power dials multiple numbers per idle agent and speeds up or slows down against a configured abandonment threshold. Predictive launches calls before the rep is free using the stage-based algorithm. Agentless uses no agents or script at all and responds to live parties and answering machines from a call analysis response set, which Genesys notes can transfer a live party to an outbound flow but not directly to an agent.

The sixth is the one worth knowing about. External calling exists, in Genesys' own words, so that the campaign uses a third-party desktop dialer for Telephone Customer Protection Act compliance. The rep reviews the record, copies the number and dials it elsewhere. Genesys lists the drawbacks itself: least efficient mode, no preview timer, and a lack of call-detail tracking and recording of calls dialed with a third-party dialer. It is a compliance escape hatch that trades away most of the reason you bought a dialer.

We went through the tiers, the licence types and the arithmetic on 15 seats in our Genesys Cloud CX pricing breakdown. For teams comparing Genesys against the other enterprise platforms rather than against a small-team dialer, the seats-versus-meters comparison lives on AI call center software.

Cloud dialer security, attestation and what you inherit

Two things decide whether your calls get answered, and only one of them is in the dialer.

The first is caller ID authentication. Under the STIR/SHAKEN framework, the attestation level attached to your calls is assigned by the originating voice service provider that signs them, not by your dialing software. A self-hosted dialer signs nothing. So if you buy DIDs from one provider and terminate through a cheaper one, you are very likely to get B or C level attestation on calls that display a number the terminating carrier cannot vouch for, and the savings on termination rates rarely survive the hit to answer rate. When you buy a cloud dialer that includes telephony, you are buying that carrier relationship along with it, which is a genuine argument for the bundled option and a genuine reason to ask a hosted VICIdial provider who actually owns the trunks.

The second is number reputation, which is separate from and downstream of attestation. Signed calls still get labelled if analytics engines see a pattern they dislike. Clean signing gets you to the starting line, it does not keep you there, and the practical playbook is on how to avoid spam likely on outbound calls and branded caller ID.

On data security, the questions worth asking a cloud vendor are narrow and answerable: where recordings are stored and for how long, whether recording storage is metered separately, who inside the vendor can access transcripts, and what happens to your data at the end of the contract. Retention is often a tier feature rather than a setting. BatchDialer, for example, keeps recordings for 90, 180 and 365 days by tier, so the answer to how long can we keep call recordings is sometimes upgrade your plan.

And whichever model you choose, the DNC obligations, the calling-window rules and the consent record stay yours. A vendor can give you the tooling. It cannot take on the liability. Our TCPA compliance software page covers what has to be in place before a cold list gets dialed at all.

Cloud based dialer comparison: which vendors publish a price

This sounds like a trivial filter and it is actually the most useful one, because a vendor that publishes list pricing has already told you the shape of the deal.

Of the seven tools in the table above, four publish a per-agent or per-seat figure you can budget from today: BatchDialer, PhoneBurner, Genesys and Five9. Two of those come with a caveat that materially changes the number. Five9's fifty-seat minimum is in a footnote on its own pricing page, and Genesys shows a Named-licence figure while offering Concurrent and Hourly Interacting licence types with no published rate at all.

Convoso publishes nothing, sells annually and does not offer a free trial. Kixie, in the adjacent sales-dialer category, renders three plan names in a browser with no figure on any of them and sells DNC scrubbing as a compliance add-on. Orum prints MINIMUM 9-SEAT above its plan and no price. Vulcan7 publishes three packages and no prices. None of that makes them bad products. It does mean the evaluation has to happen inside a sales cycle, and that you will not know whether they are affordable until you have spent two weeks finding out.

Our own row is on the same table and it says planned launch pricing, waitlist only, because that is the truth today. What differs is the shape rather than the number: an AI voice agent that holds the conversation has no agent queue, so there is no concurrency licence to size, no fifteen-agent floor before the pacing model works, and no abandonment rate to manage against a three percent budget. That is a different product to a dialer, and if you want the honest side-by-side of both approaches it is on best AI cold calling software and AI dialer.