The Real Cost Behind "Cheap" Per-Minute Rates

You've seen the ads: AI calling agents for $0.05 a minute. Here's the uncomfortable truth — that number is rarely what you actually pay, and the gap between the two is where most budgets go to die.

The problem has a name: BYOK, or "bring your own keys." Many platforms advertise a low orchestration rate, then require you to pay OpenAI, ElevenLabs, Deepgram, and Twilio separately for the actual intelligence and phone lines. One pricing analysis found this hidden stack adds $0.06–$0.19 per minute, turning a "$0.05/min" platform into a $0.12–$0.25/min reality. As that analysis bluntly puts it: "If this price doesn't explicitly include the LLM, Text-to-Speech generation, and transcription, it's not a real price. It's merely an orchestration fee with expensive follow-up costs."

Even the quoted rate can hide surprises. Retell's own research shows reasoning tokens can carry a 4x premium, adding 15–25% to sophisticated agent costs. And buying guides warn of setup fees, integration fees, training fees, custom voice fees, and overage charges stacking on top.

Comparing vendors gets harder still, because the market uses four different pricing models:

• Per-minute — typical range of £0.05–£0.15/minute, but often excluding telephony and AI costs
• Per-call — £0.03–£1.00 per resolved call, which shifts accuracy risk to the vendor
• Per-seat subscription — £200–£1,000/month, pricing the software rather than the usage
• Blended pricing — a mix that makes direct comparison nearly impossible

This is why every "cheapest" claim deserves skepticism. Retell's comparison crowns Retell at $0.07/min. Famulor's comparison crowns Famulor at $0.05–$0.16/min all-inclusive — and puts Retell's true cost at $0.13–$0.24/min. Both are vendor self-assessments, not independent audits. No neutral pricing study exists in this space; every leaderboard you see was published by someone competing on it.

The deeper issue is that cents-per-minute measures the wrong thing. An agent that quotes the wrong price or invents a policy does more damage than a voicemail, because the customer believes it. That's why Worqd prices AI calling work against outcomes — qualified conversations and booked calls — rather than chasing the lowest headline rate, and why the cheapest agent on paper is rarely the cheapest in practice.

The Cheapest Options, Compared Honestly

Every "cheapest AI calling agent" comparison you'll find online was written by the company that won. That doesn't make the numbers useless — but it means you should read them with one hand on your wallet.

The low-cost leaders, with attribution

Three names consistently appear at the bottom of the pricing table. Retell AI claims $0.07/min, dropping to roughly $0.05/min at enterprise scale, with a 10,000-minute bill around $700. Famulor advertises $0.05–$0.16/min all-inclusive, starting at $0/month. And Grok Voice lists $0.08/minute for speech-to-speech agents, alongside a claimed 67.3% score on the τ-voice Bench Leaderboard.

Advertised vs. true cost, side by side

Here's where honesty matters. Those sticker prices often exclude the components that make a call actually work — the language model, text-to-speech, transcription, and phone lines. According to Famulor's own pricing breakdown, the "BYOK trap" — bring your own keys — inflates advertised prices by $0.06–$0.19/min.

• Vapi: $0.05/min advertised, but $0.12–$0.25/min true cost
• Bland AI: $0.07/min advertised, $0.10–$0.18/min true cost
• Synthflow: $29/mo base, but $0.15–$0.27/min true cost
• Retell AI: $0.07/min claimed by Retell — but Famulor puts its true cost at $0.13–$0.24/min

As one analysis puts it: "If this price doesn't explicitly include the LLM, Text-to-Speech generation, and transcription, it's not a real price." Every source crowning a "cheapest" option ranks itself first — treat those claims as promotional.

Why cents-per-minute is the wrong question

Zoom out and the per-minute differences almost disappear. One cost comparison puts AI calling at $0.30–$0.50 per call versus $6.00–$7.68 for a human call center — a 93–95% reduction. At 60,000 calls a year, that's $18,000–$30,000 versus $360,000–$460,000.

When the gap between AI and human is that large, arguing over $0.02/min between platforms misses the point. What matters is whether the call converts: whether the agent books the appointment, qualifies the lead, or hands off to a person with full context. Experts warn that the biggest risk isn't overpaying — it's a cheap agent that confidently quotes the wrong price or invents a policy.

This is exactly why Worqd scopes its AI calling work around outcomes — booked calls, qualified conversations — rather than competing on the lowest per-minute rate. The cheapest minute that never books a call costs you everything.

Why Cheap Minutes Can Cost You Leads

A $0.05-per-minute rate looks irresistible — until the agent quotes the wrong price to a hot lead. The cheapest AI calling agents save you cents on every call, but what happens on those calls determines whether you actually grow.

The core problem is what voice AI experts call the biggest risk in this space: confident wrong answers. An agent that invents a policy or gives the wrong hours does more damage than a voicemail, because the customer believes it. A missed call is a lost opportunity. A wrong answer is a damaged relationship.

The pricing itself is often an illusion. Independent cost analysis shows that platforms advertising $0.05/min often cost $0.12–$0.25/min once you add separate bills for the language model, speech-to-text, and text-to-speech — the "bring your own keys" trap adds $0.06–$0.19/min to the sticker price. As that same analysis puts it, if a price doesn't include those pieces, "it's merely an orchestration fee with expensive follow-up costs."

Then there's the second hidden cost: your time. The market splits between self-serve platforms and done-for-you providers, and the cheap end is almost always self-serve. That means you own:

• Building and maintaining the agent yourself, across multiple vendor accounts
• Connecting it to your calendar, CRM, and follow-up processes
• Monitoring calls for the wrong answers that quietly burn leads
• Troubleshooting when the pieces stop talking to each other

As one pricing guide bluntly notes, "cheaper with 4 accounts and manual troubleshooting" is an administrative nightmare for a small business, not a win. And buying guidance is clear: most businesses without in-house AI teams benefit more from a provider who builds the agent and stays involved.

This is why Worqd deliberately prices against results — booked calls and qualified conversations — rather than competing on cents per minute. A cheap agent that can't touch your calendar or CRM is, in the words of one expert, "a glorified answering machine." The value isn't in the minutes. It's in what the call produces: the appointment booked, the lead logged, the follow-up text sent — with a real person ready to take over when the conversation gets serious.

What to Look For Instead of the Lowest Price

The cheapest AI calling agent on paper is rarely the cheapest in practice. Advertised rates of $0.05 per minute often balloon to $0.12–$0.25 once you add the separate bills for speech models, transcription, and phone lines — the so-called "BYOK trap" that adds $0.06–$0.19 per minute to platforms that look cheap upfront, according to one detailed pricing analysis. As another buyer's guide puts it: "Two agents can have very different sticker prices and very similar real costs, or the reverse."

So if cents-per-minute is the wrong yardstick, what should you measure instead? Four things separate an agent that earns its keep from one that quietly costs you leads.

• Accuracy under real conditions. A cheap agent that invents a policy or quotes the wrong price does more damage than a voicemail, because the customer believes it.
• Integration with your CRM and calendar. An agent that can't touch your other systems is, in the words of one guide, "a glorified answering machine."
• Human handoff with full context, so a warm lead never has to repeat themselves.
• Speed-to-booked-call — how fast an inquiry becomes a confirmed appointment on your calendar.

The stakes are bigger than they look. Voice AI handles calls for $0.30–$0.50 versus $6.00–$7.68 at a human call center — a 93–95% reduction, per this cost comparison. When the baseline is that cheap, shaving two cents per minute saves almost nothing while a missed or mishandled call costs a real customer.

There's also a structural divide in the market: self-serve platforms you build and maintain yourself, versus providers who build it with you and stay involved. Most businesses without in-house AI teams do better with the second approach — "cheaper with 4 accounts and manual troubleshooting is not a win for a small business," as one analysis bluntly notes.

That's why Worqd deliberately stays out of the per-minute price war. Every inquiry gets qualified in under 60 seconds, 24/7 — including nights and weekends — with calls handed to a real person carrying full context, booked straight into your calendar and rules. And because the work is priced against the results that matter rather than minutes logged, the incentive lines up with yours: more booked calls, not longer calls.

How to Choose the Right Option for Your Business

The cheapest sticker price rarely wins. By the time you add up the real costs, the "budget" option often costs more than the premium one — and performs worse when it matters.

Calculate the true total cost, not the sticker price. Advertised rates of $0.05/minute can climb to $0.12–$0.25 once you add the separate bills for the LLM, speech-to-text, and text-to-speech — what one pricing analysis calls "an orchestration fee with expensive follow-up costs." Also watch for setup, integration, training, and overage fees, which buying guides flag as the most common budget surprises.

Test accuracy before you commit. The biggest risk with a cheap agent isn't the price — it's confident wrong answers. As one expert guide puts it, an agent that quotes the wrong price or invents a policy "does more damage than a voicemail, because the customer believes it." Run real-world calls, not polished demos, before signing anything.

Check integration and handoff capability. A voice agent that can't touch your calendar, CRM, and follow-up systems is, in the words of that same guide, "a glorified answering machine." The value comes from what happens after the call: the booked appointment, the logged lead, the text that goes out. Make sure calls can hand off to a human with full context when the conversation gets complicated.

Decide whether you can actually run it yourself. The market splits between self-serve platforms you build and maintain, and providers who build it for you and stay involved. As one pricing comparison bluntly notes, "cheaper with 4 accounts and manual troubleshooting is not a win for a small business — it's an administrative nightmare." Ask yourself honestly:

• Who on your team will build, test, and maintain the agent week after week?
• Who fixes it at 9 p.m. on a Friday when a call goes wrong?
• Who monitors whether the agent is actually booking calls — or just answering them?

If you don't have clear answers, a managed approach fits better. That's the tradeoff at the heart of this decision: integrated beats fragmented. One partner running the whole path from first click to booked call, with one plan and one report, avoids the vendor-juggling that erodes both savings and results. Remember the benchmark: AI calling runs $0.30–$0.50 per call versus $6.00–$7.68 for humans — so the real question isn't "what's the cheapest agent?" It's "which option turns more of your leads into revenue?"

That's why Worqd scopes every engagement on a free growth call rather than publishing per-minute rates — the right fit depends on your call volume, your systems, and what a booked call is worth to you. Book a growth call and we'll help you find the bottleneck before you spend a dollar on any agent, cheap or otherwise.