Product · Billing

How much does 1 AI call cost? Real breakdown of 6 factors

In marketing we say ~50 credits ≈ 1 call. That's the average. In reality a single call can cost 30 to 90 credits. We show 6 factors driving price, how to control them, and when to consciously accept a more expensive call.

CallBotAgent Team30 Apr 2026~10 min readProduct · Billing
Circular gauge dashboard with needle at '50' (highlighted emerald), with mini-dials around: LLM, TTS, KB, Script, Add-ons, Sec.
TL;DR

AI call cost is the sum of 6 dynamic per-second components. Average ~50 credits for a 113-second confirmation call with standard model. Cheap profile: 30 cr (short call, basic model, no add-ons). Expensive profile: 90 cr (long call, premium model, KB + transcription). No hidden mark-up, no rounding-up.

  • ~50 credits ≈ 1 call is average, not literal fixed rate.
  • Real spread: 30-90 credits depending on 6 factors.
  • We bill per-second, no rounding-up like competitors.
  • Full breakdown in Pricing Calculator + 20-min call option.
6 factors

What drives cost: 6 dynamic levers

Billing model is per-second usage. Each second has a variable price depending on call profile. Mathematically simple, but requires understanding profile composition.

Six cost levers: (1) chosen LLM model, (2) knowledge base length attached to script, (3) script length itself, (4) TTS engine (text-to-speech), (5) add-ons (PII redact, denoising, guardrails, QA), (6) actual call duration in seconds.

  • 1LLM model (GPT 4.1 / GPT 5 / Claude 4.5 / Gemini). Different $/min rates. Standard production mix = baseline. Premium GPT 5 = +30% cost.
  • 2Connected knowledge base. Each KB attached = +$0.005/min. Store with 5 KBs pays +$0.025/min vs baseline.
  • 3Script length. More tokens = more cost. 5-page script = baseline. 15-page = +20%.
  • 4TTS engine. Premium ~$0.040/min vs standard ~$0.015/min. Most pilots use standard.
  • 5Add-ons. PII redact +$0.01/min · denoising/guardrails +$0.005/min · QA +$0.10/min after first 100 min/mo.
  • 6Actual call duration. Per-second billing means: 90-second call literally costs 90/113 less than 113-second average.
Hexagonal honeycomb of 6 cost factors: LLM model, TTS, Knowledge base, Script length, Add-ons, Per-second
Real examples

Three call profiles: 30 / 50 / 90 credits

Cheap profile (30 cr): 75-second confirmation, GPT 4.1 standard, 3-page script, standard TTS, no add-ons, no KB. Typical for small categories (single-SKU cosmetics, single-product orders).

Average profile (50 cr): 113-second call, standard model, 5-page script + 1 KB (FAQ), standard TTS, denoising. Typical for most pilots.

Expensive profile (90 cr): 145-second call (health/wellness needs longer explanations), GPT 5, 8-page script + 3 KBs (catalog + FAQ + compliance), premium TTS, PII redact + QA. Typical for Scale+ with full transcription and analytics.

Calculator + custom estimate

Calculate for your store in the Calculator — or book 20 min

Simplest path: Pricing Calculator takes call volume × average length × chosen plan and returns monthly estimate. Linear from call count × ~50 cr/call.

If profile is unusual (long wellness calls, multilingual, full QA suite), book a 20-minute call with our team — we'll calculate custom estimate based on your script and volume. Book a call →

Calculator UI mockup: 3 sliders (calls, length, add-ons) + violet result card 'Estimate: 2,145 cr/mo' + Plan Growth OK
What sets us apart

Per-second, no rounding-up

Most competitor voice AI providers bill per minute with rounding-up. A 65-second call costs as much as 120 seconds (rounded up to second full minute). At 1000 calls/mo that's ~12% cost difference.

We do per-second billing: 65s = 65/60 minute. No hidden mark-up. No surprise invoice. Full billing transparency in panel with per-call breakdown (Scale+) or per-day (Growth+).

If you notice calls in panel much pricier than average (above 90 cr), one click shows breakdown per factor — decide what to optimize: shorter script, fewer KBs, standard TTS instead of premium.

Conclusions

Five things to remember

1

~50 cr ≈ 1 call is average, not fixed. Spread 30-90 cr depends on 6 factors. Marketing simple, billing precise.

2

Per-second, no rounding-up. 65-second call = 65/60 min. With competitors it's 2 full minutes.

3

Biggest levers: LLM model, call length, KB. Optimization starts with shorter script (easiest win).

4

Add-ons are optional and visible in pricing. PII redact, denoising, QA — each separate rate. No hidden premium.

5

Custom estimate for unusual profiles. 20-min call with team + breakdown based on your script and volume.

7-day trial with real credits

Real call cost?Test 1,000 credits in 7 days.

1,000 trial credits (~20 confirmation calls) · full access to your chosen plan · cancel anytime. Card required · no charge during trial · auto-activates on day 8.

Cancel anytimeEmail support includedGDPR · EU data residencyCard required · $0 during trial