An AI-assisted first interview typically costs a low single-digit euro amount per conversation, depending on the vendor and billing model - clearly less than the staff time a manually run interview requires. The exact figure depends heavily on the billing model: usage-based via credits, a flat per-seat fee, or bundled into a broader applicant-tracking package.
For companies trying to size the real cost, comparing the actual billing logic matters more than chasing a single list price - that is exactly where vendors differ the most.
The three common billing models
| Model | How billing works | Good fit for |
|---|---|---|
| Usage-based (credits/per conversation) | Cost only for conversations actually held | Fluctuating or seasonal application volume |
| Flat per-seat licence | Fixed monthly fee per user regardless of volume | Consistently high, predictable volume |
| Bundled into a broader package | Interview feature included in the price of a wider ATS/recruiting suite | Companies already rolling out a new full system |
Usage-based models are usually more predictable for companies with fluctuating application volume: cost is only incurred when a conversation actually happens, rather than paying a fixed base fee regardless of volume.
A worked example of usage-based billing
At Atlas, usage is billed in credits. A credit typically costs around 7 cents, depending on the package booked. A 5-minute AI voice interview consumes 28 credits - at that credit price, roughly 1.96 euros per conversation. A full AI CV-screening assessment consumes 3 credits, or roughly 21 cents.
Scaled up, the difference from manual handling becomes clearer: 100 first interviews run as 5-minute voice interviews cost roughly 196 euros on this basis. Running the same number of first interviews manually typically requires substantially more staff hours once scheduling, conducting the conversation, and writing it up are realistically accounted for.
One important caveat for this comparison: it does not replace human judgment or the decisive, later conversation with the shortlist. It replaces the groundwork - the first, structured screening conversation that nobody manually delivers at this scale today, simply because there is not enough time.
What actually drives the price per conversation
- Conversation length. A longer, more in-depth interview consumes more compute than a short pre-screen and costs more accordingly.
- Language and channel. A phone call or a WhatsApp voice message can carry different costs than a browser-based conversation, depending on the underlying infrastructure.
- Post-processing. Transcription, summarisation and structured evaluation of the answers are bundled into the base price by some vendors and billed separately by others.
- Volume. As with most usage-based models, the per-unit price often drops as total volume increases.
What a pure price comparison often misses
A low price per conversation says nothing about how many of those conversations actually lead to a usable assessment. A cheap but shallow interview that still requires an extra manual screening afterwards is saving in the wrong place. What matters more than the raw unit price is how much of the process - from scheduling through to structured evaluation - is already included in it.
A checklist for comparing vendors
A pure price comparison between vendors rarely leads to a solid decision, because service descriptions are rarely equivalent. It works better to ask the same questions of every offer:
- Is the price calculated per conversation or per seat - and does that match actual application volume?
- Are transcription and structured evaluation included in the price, or billed separately?
- Do extra costs apply for different languages or channels (phone, WhatsApp, browser)?
- Is there a minimum commitment or contract term that becomes a disadvantage with fluctuating volume?
- How transparently does the vendor disclose the criteria a conversation is scored against?
Answering these questions for two or three offers in parallel gives a far more solid basis for comparison than a single quoted price per interview.
Hidden costs that are easy to miss
Beyond the raw per-conversation price, it is worth looking at three cost items that rarely show up immediately in a quote: setting up the system (integration with the existing applicant-tracking system), maintaining interview questions and scoring criteria over time, and the internal effort of reviewing AI outputs - especially early on, while trust in the scoring is still being built. These line items rarely appear on a price sheet, but they have a real effect on total cost. Raising them openly before signing avoids unpleasant surprises afterwards.
Frequently asked questions about AI interview costs
Are there free AI interview tools?
Some vendors offer a free trial period or a limited number of free conversations, but a permanently free full deployment at meaningful volume is the exception in the market, because every conversation involves real compute and speech-processing cost.
Is an AI interview cheaper than a traditional phone screen?
On a pure unit-cost basis, usually yes, because no staff time is spent on the conversation itself. The real advantage, though, often lies less in the per-unit price than in being able to run far more conversations in the same amount of time without scaling recruiting headcount proportionally.
Are there extra costs for evaluation?
That depends on the vendor. Some include transcription and structured summarisation in the base price per conversation, others bill it separately - worth checking closely in the service description before signing.
Does it make sense at low application volume too?
With a usage-based model, cost is only incurred for conversations actually held - which makes it economically viable at lower volume too, unlike a fixed monthly base fee charged regardless of usage.
Does an AI interview fully replace the in-person conversation?
No. An AI-assisted first interview handles structured pre-screening and information gathering - the final decision and getting to know the shortlist in person remain tasks for the humans in the process.
Is per-seat or per-conversation billing more common?
Both models coexist in the market, without a clear standard having emerged. Larger, established recruiting suites tend towards per-seat licences, while specialised AI interview vendors more often use usage-based models - checking the specific vendor's pricing page remains necessary either way.
Does the price change with multiple languages?
That depends on the underlying language model and speech recognition used. Some vendors charge the same price per conversation across all supported languages, others tier pricing by language or region - worth asking about explicitly before signing, especially with an internationally diverse applicant pool.
Details on how an AI voice interview actually works are on the Atlas voice interview page, with more articles in the AI interviews topic hub. The legal side of using AI in interviews is covered in Is AI Emotion Recognition Allowed in Job Interviews?, and an overview of common vendors is in the AI interview & voice tools directory.
