"The cheapest research is the research you actually do — consistently, at scale, without the invoice shock."

When research teams evaluate moving from human telephone interviewing to AI voice, the first question is always the same: what will it cost?

It's the right question. But it's rarely answered completely. Vendor conversations tend to zoom in on the per-minute or per-interview rate while quietly glossing over all the costs that don't appear on a platform invoice — the ones that make human telephony far more expensive than it looks on a briefing document.

This article breaks it all down. We'll build a full, honest cost model for a human telephony operation, compare it to an AI voice survey platform, and show you what the numbers actually look like on a real study.

The Full Cost of a Human Telephony Interview

Most organisations budget for telephone interviewing by looking at the rate card from their fieldwork supplier — a cost-per-completed-interview figure that looks reasonable in isolation. What that figure bundles, and what it conceals, is where the real story begins.

Labour: the dominant cost driver

A human telephone interview requires a trained interviewer on the line for every single minute of every call. Depending on market, seniority, and whether the work is handled in-house or by an agency, interviewer labour typically runs between $8 and $18 per hour. For a 10-minute interview that takes 20 minutes of clock time (accounting for dials, rings, refusals, and introductions before reaching a cooperative respondent), you are spending roughly $3 to $6 in interviewer time per completed interview before a single data point has been recorded.

Scale that to a 1,000-respondent study and you already have a labour line between $3,000 and $6,000 — just for the interviewers.

Management overhead

Call centre operations don't run themselves. Supervisors monitor call quality, manage shift schedules, handle respondent escalations, and run daily briefings. Industry benchmarks put supervision ratios at roughly 1 supervisor per 8–12 interviewers. For a mid-sized study, that adds 10–15% on top of the base interviewer cost.

Add project management: a fieldwork coordinator tracking quotas, managing the sample, and liaising with the research team. On a standard two-week field period, that's 3–5 days of coordinator time.

Retry logic — the hidden multiplier

Not every dialled number reaches a completed interview. A typical human telephony operation needs to dial 3 to 6 numbers for every completed interview it produces, accounting for non-contacts, refusals, engaged lines, voicemails, and ineligible respondents. Every one of those dials costs interviewer time, even when nothing happens.

This is the multiplier most budgets underestimate. If your cost-per-dial is $0.80 and you need 5 dials per completion, that's $4.00 of dial cost per interview — before you've actually spoken to anyone.

Quality assurance and back-checks

Research industry standards (and client contract obligations) typically require a percentage of completed interviews to be back-checked — a supervisor calls the respondent back to verify the interview took place and a sample of questions were answered correctly. Industry standard is 10–20% back-check rate.

Each back-check costs roughly the same as the original interview. At 15% back-check rate on a 1,000-interview study, that's 150 additional calls — a material cost line that rarely appears in initial project quotes.

The total picture

When you add together interviewer labour, supervision, project management, retry dials, back-checks, and the platform or software cost underlying the CATI system itself, the true all-in cost of a human telephony interview lands between $9.80 and $21.00 per completed interview.

For a 1,000-respondent study in a single market, that's $9,800 to $21,000 in fieldwork cost alone — before analysis, reporting, or any platform subscription. For a multinational study requiring two or three language versions? Multiply accordingly.

What AI Voice Actually Costs

AI voice survey platforms are priced differently — and more transparently. The dominant model is per-minute billing, where you pay for the duration of actual voice calls conducted. There are no retry labour costs (failed dials cost nothing), no supervision overhead, no back-check budget, and no management hours to allocate.

Using Voiceter.ai's published pricing as a reference point: Pay As You Go charges $0.45/min (Native) with no monthly fee. Silver is $349/mo at $0.34/min. Gold is $749/mo at $0.30/min. Platinum is $1,999/mo at $0.267/min. Agency is $2,999/mo at $0.15/min.

For a 10-minute interview, even at the entry-level Pay As You Go rate, the cost is $4.50 per completed interview — and that includes transcription, real-time sentiment analysis, audio quality scoring, and a full analytics layer. There is no additional line for any of those.

The Side-by-Side: A 1,000-Respondent Study

Let's model a concrete study: 1,000 completed interviews, 10-minute average interview length, single market, single language.

Human telephony total: $11,800–$24,500 ($11.80–$24.50 per interview). Voiceter AI on the Silver plan: $3,749 total — $3,400 in voice minutes (1,000 × 10 min × $0.34) plus $349 monthly subscription. Cost per completed interview: $3.75.

The AI voice option delivers the same 1,000 interviews for roughly 70–84% less than a comparable human telephony operation — and produces richer outputs, because every call is automatically transcribed, sentiment-scored, and fed into an analytics dashboard in real time.

"But What About Quality?"

This is the objection every research buyer eventually raises, and it deserves a direct answer.

The consistency argument actually favours AI. A human interviewer on their eighth hour of calls is not performing identically to the same interviewer at the start of their shift. Tone, pacing, slight variations in question delivery, unconscious probing — these are documented sources of interviewer effect in survey methodology literature. An AI voice agent is identical on call 1 and call 10,000.

What AI voice does not yet replicate is the deeply qualitative dimension of a skilled human interviewer — the intuition to probe an unexpected response, the social intelligence to navigate a sensitive moment in a conversation. For exploratory qualitative research and executive interviews, human interviewers remain the right tool. But for the large-volume quantitative fieldwork that makes up the majority of telephone survey programmes — brand tracking, customer satisfaction, public opinion — AI voice is not a quality compromise. It is a quality upgrade in the dimensions that matter most at scale: consistency, completeness, and speed.

How to Model This for Your Own Programme

Different research programmes have different interview lengths, completion targets, and frequency. Here's a simple framework to calculate your own comparison:

Human telephony cost: (Interviews needed × average interview length in minutes × $1.20–2.40 per minute all-in) + fixed management overhead.

Voiceter AI cost: (Monthly plan fee) + (interviews needed × average interview length × per-minute rate).

At Silver plan rates ($0.34/min), a 5-minute interview costs $1.70. At Platinum rates ($0.267/min) for programmes running monthly tracking studies, the cost drops below $1.35 per completed interview — including full analytics.

For agencies running multiple concurrent client projects, the Agency plan ($2,999/mo at $0.15/min) delivers a per-interview cost of $0.75 on a 5-minute study — a figure that simply cannot be achieved through human telephony at any scale.

How to Present This Internally

If you're building a business case to move your telephone research programme to AI voice, here's the framing that lands best with finance and procurement stakeholders:

• Lead with total cost of ownership, not rate cards — the all-in cost including management overhead, retries, QA, and platform fees makes the AI option decisively cheaper
• Quantify the speed benefit in budget terms — a study that takes 3 weeks to field with a human call centre and 3 days with AI voice frees 2.5 weeks of project management time
• Frame the analytics as included, not extra — real-time sentiment analysis, 100% transcription, and automated executive summaries are built in, not additional budget lines

A Final Note on Transparency

We've published our pricing openly on the Voiceter.ai pricing page because we believe research buyers deserve to model costs accurately before a single sales conversation. The numbers in this article reflect our published rates; we're not hiding overages or per-seat fees in the footnotes.

If you want to run the model against your actual programme — interview volume, average length, number of markets, frequency — our team is happy to build it with you.