Why Advertised Per-Minute Rates Mislead

Headline rates like 9¢ per connected minute tell only part of the story, as they represent just 30-50% of the true cost per hour for enterprise calling campaigns. This significant gap arises because advertised per-minute rates fail to capture the full cost structure that buyers actually encounter on their invoices.

Three compounding multiplier layers drive this underestimation: the base rate itself, the billable minute definition (which includes hold time, silence, IVR prompts, and ring time), and failure/transfer/spike penalties such as concurrency ceilings and compliance surcharges. These factors multiply rather than add, causing enterprise buyers to consistently underestimate true costs by 40-60% when auditing base rates alone.

To illustrate, the billable minute expansion factor typically ranges from 1.0x to 1.5x or higher, meaning a 60-second call may be billed as 60 to 90+ seconds depending on how vendors define billable time. Additionally, concurrency penalties apply when call volume exceeds platform limits, with overage rates often 30-50% above the base per-minute rate. These layered multipliers compound quickly—turning what seems like a low per-minute rate into a substantially higher effective cost per hour.

For My AI Call Center campaigns, this means the starting rate of 9¢ per connected minute must be evaluated within this broader context of billable definitions and potential overage charges to avoid budget surprises. Understanding these mechanics is essential for accurate call cost per hour calculations and informed campaign planning.

The Full Cost Stack: Breaking Down Every Layer

Every AI call minute you pay for is actually five bills stacked on top of each other — and buyers who only look at the headline rate routinely miss 40-60% of their true cost, according to pricing analysis from Bland.ai. Understanding each layer is the difference between an accurate cost-per-hour calculation and an invoice surprise.

Layer one: telephony. Carrying the call itself costs $0.005-$0.022 per minute. Twilio charges $0.0085/min for inbound local and $0.022/min for toll-free, while Telnyx routes through Vapi at $0.0055/min, per cost breakdowns from Zenthos. Outbound calling sits in the middle at roughly $0.014/min on Twilio.

Layer two: speech-to-text. Converting the caller's voice to text runs $0.005-$0.01/min. Deepgram's Nova-3 model comes in at $0.0048-$0.0058/min, while OpenAI's real-time transcription runs $0.017/min — and streaming STT costs 50-80% more than batch processing.

Layer three: the LLM. This is where budgets blow up, because the range is a 50x spread: from $0.003/min on GPT-5 nano to $0.16/min on frontier models. Model choice alone can swing your hourly cost by more than $9.

Layer four: text-to-speech. Neural voices cost $0.012-$0.04/min — Deepgram Aura-2 at the low end, ElevenLabs Flash around $0.021, per the Zenthos component analysis.

Layer five: the platform fee. Vapi charges $0.05/min, Retell $0.055, and ElevenLabs Agents $0.08 — the toll for orchestrating everything.

Stack those layers and the totals emerge: managed platforms land at $0.07-$0.31/min all-in, while self-built stacks run $0.04-$0.10/min — but you supply the engineering. Then the hidden costs arrive, adding another 20-40% to first bills:

• Setup and build fees ranging from zero to several thousand dollars
• Phone numbers at $2-$15/month each
• Concurrent line fees of $8-$10 per line, per month
• Feature add-ons like booking, SMS, and bilingual support

This is why My AI Call Center quotes the whole campaign before launch — one rate starting at 9¢ per connected minute, locked for the campaign, with no per-seat charges or platform bill hiding underneath. The number you approve is the number you pay. When comparing vendors, ask which of these five layers their rate actually includes — because platforms sell raw minutes while managed services sell finished outcomes, and only one of those numbers predicts your invoice.

From Cost Per Hour to Cost Per Resolved Call

A vendor quotes you $0.10 per minute, and the slide deck says your AI calls cost $0.60 each. Then the invoice arrives, and the real number is more than double. The gap isn't a billing error — it's what happens when you measure the wrong metric.

Per-minute and per-hour rates tell you what you spend while the clock runs. They don't tell you what you get. As one industry analysis puts it plainly: "Cost per resolved call is the only number that tells you whether AI saves money." When a cheap AI minute fails and the call goes to a human, you pay for both.

The formula for cost per resolved call is:

(AI cost per minute × average call duration) + (human cost per minute × average handoff duration × (1 − containment rate))

The containment rate is the share of calls the AI completes without escalation. The rest transfer to a human, whose time you also pay for — and fully loaded U.S. human agents run $25–$45 per hour once benefits are included.

Here's the worked example that surprises most buyers. Take an AI stack at $0.15 per minute all-in, 4-minute average calls, and offshore human backup at $0.25 per minute. At 30% containment, your AI spend per actually-resolved call hits $1.30 — more than double the $0.60 vendor-slide figure, according to the same analysis. At that containment level, the AI-plus-human mix actually costs more per month than an all-human offshore team.

Push containment to 50% and the picture flips: the blended cost drops to $0.95 per call. At 70% containment against onshore humans, savings reach roughly $92,000 per month for a 40-seat operation. Same per-minute rate, wildly different economics.

Two levers control everything:

• Containment rate — every failed AI call adds a human minute bill on top of the AI minutes you already paid for.
• Average call duration — longer calls multiply your per-minute rate, and billable-minute definitions (hold time, silence, ring time) can expand billed minutes by 1.0x–1.5x, per Bland's pricing analysis.
• Handoff duration — how long the human spends rescuing an escalated call.

This is why campaign design matters more than headline rate. A structured campaign with one clear goal per call — confirm, qualify, remind — keeps calls short and containment high. That's the thinking behind My AI Call Center's campaign model: outcomes are disposition-coded and reported per call, so you can compute cost per resolved call from actual results rather than vendor estimates. Managed outbound campaigns start at 9¢ per connected minute, quoted before launch with the rate locked for the campaign.

Run the math on resolved calls, not connected minutes. It's the only calculation that tells you whether AI actually saves money — before you spend it.

Benchmarking Against Human Agent Costs by Region

When evaluating call center solutions, understanding how AI compares to human labor costs across regions is essential for accurate budgeting. Research shows fully loaded human agent costs vary significantly: U.S./Canada operations range from $28-$42 per hour, Western Europe from $30-$41, Eastern Europe from $12-$22, Latin America from $12-$24, the Philippines from $7-$16, and India from $5-$14 per agent hour. These figures include base wages and mandatory benefits, with U.S. private industry benefits adding 45.6% on top of base salary according to BLS ECEC March 2026 data. For My AI Call Center’s Halifax and Austin operations, the U.S./Canada benchmark provides the most relevant comparison point for onshore evaluations.

To enable apples-to-apples cost analysis, these hourly rates can be converted to productive-minute costs based on typical agent utilization. Industry calculations show that with 48 productive minutes per hour (accounting for breaks, training, and system downtime), onshore human labor costs approximately $0.73 per minute ($35/hr ÷ 48 productive min), while offshore equivalents average $0.25 per minute ($12/hr ÷ 48). This conversion allows direct comparison with AI voice agent pricing, which My AI Call Center offers starting at 9¢ per connected minute for managed campaigns. When assessing true cost efficiency, businesses must consider not just per-minute rates but also billable minute expansion factors—typically 1.0x to 1.5x+—which account for hold time, silence, IVR prompts, and ring time that vendors often exclude from base quotes.

Hybrid delivery models further optimize economics by leveraging AI for high-volume routine interactions while reserving human agents for complex or sensitive cases. Research indicates that 60-70% of inbound calls are routine and suitable for automation, allowing organizations to capture most AI-driven savings while maintaining human oversight for scenarios requiring empathy or judgment. In such models, AI handles the majority of volume at a fraction of human cost, with agents focusing on escalations, relationship-building, or exception handling. This approach reduces overall cost per resolved call without sacrificing service quality, particularly when AI containment rates exceed 50%. By grounding comparisons in verified regional benchmarks and productive-hour calculations, decision-makers can accurately assess whether My AI Call Center’s managed AI calling campaigns deliver measurable cost advantages over traditional staffing models for their specific use cases and geographic footprint.

Applying the Framework to Your Campaign

Applying the calculation framework to your campaign starts with building a complete cost picture before launch. My AI Call Center’s managed service model requires summing all cost stack layers—telephony, STT, LLM, TTS, and platform fees—to determine an effective per-minute rate, which research shows typically represents only 30-50% of actual costs when hidden variables are ignored according to industry analysis. This transparent approach aligns with the "no invented numbers" promise by making every variable visible before any spend occurs.

Next, apply the billable minute expansion factor based on your campaign type—reminders, qualification, or surveys—since hold time, silence, and IVR prompts can inflate bills by up to 50% at identical headline rates as demonstrated in pricing studies. For example, qualification campaigns often run longer with more conversational turns, increasing the expansion factor toward 1.5x, while simple reminders may stay closer to 1.0x. Then layer in platform fees and concurrency buffers, noting that overage rates frequently run 30-50% above base during seasonal spikes per vendor billing practices.

Finally, calculate cost per resolved outcome using campaign-specific containment expectations. Since AI handles only the routine portion of interactions (60-70% per Gartner), the true cost per confirmed appointment or qualified lead must account for any handoff to human agents per outsourcing benchmarks. This outcome-focused metric—cost per resolved call—better reflects campaign effectiveness than hourly rates alone, especially when containment rates fall below 100%. By walking through each layer—cost stack, expansion factors, fees, and outcome efficiency—you gain a launch-ready worksheet that prevents invoice shock and supports accurate ROI modeling for clinics, franchises, and membership businesses alike.