Conversational AI in hiring
Conversational AI is the technology that lets software hold a spoken or written conversation in everyday language. It combines speech recognition, a language model and speech synthesis. In hiring it lets an AI assistant call or message candidates, understand their answers and respond naturally.
Conversational AI is the technology underneath. Conversational screening is one way to use it: screening candidates through a conversation instead of a form. An AI agent is the software that uses it to do a task on its own. This page is about the technology itself.
What is conversational AI built from?
A spoken conversation runs through three parts, many times per minute:
- Speech recognition. The candidate's voice is turned into text.
- A language model. The software reads the text, understands what the candidate means and decides what to say next.
- Speech synthesis. The answer is turned back into a natural-sounding voice.
In a chat such as WhatsApp the first and last steps fall away, because the conversation is already text. Voice messages in a chat go through speech recognition first.
Around these three parts sit the rules of the job: the questions to ask, what counts as a good answer and when to hand over to a person.
How does conversational AI differ from a phone menu?
| Phone menu (IVR) | Conversational AI | |
|---|---|---|
| How the caller answers | Presses a key or says one fixed word | Talks in their own words |
| What it understands | Only the options in the menu | Free answers, including ones it did not expect |
| Questions from the caller | Not possible | The caller can ask and get an answer |
| Who starts the call | Usually the caller | The software can call the candidate |
| How it feels | Like a machine | Closer to a conversation with a person |
A phone menu works for routing a call to the right desk. A screening needs answers in the candidate's own words, which a menu cannot take.
What can go wrong?
- Mishearing. Names, places and job titles are easy to get wrong. Good systems learn the words that appear in the vacancy.
- Background noise. Candidates call from a warehouse floor or a bus. Recognition gets harder.
- Inventing answers. A language model can sound sure about something it does not know. It should stick to the facts it has been given and hand over when it is unsure.
- Going too far. Some tools try to read emotions or stress from a voice. In EU hiring that has been banned since 2 February 2025. See EU AI Act in recruitment.
Why does it suit frontline hiring?
Frontline candidates rarely fill in long forms and often do not read email. They do pick up the phone in the evening and answer WhatsApp in a break. Conversational AI lets a hiring team meet them there, at any volume and at any hour, in a way that feels like talking to a person.
How does Ringtime use conversational AI?
Ringtime's AI agents use conversational AI to screen candidates by phone and WhatsApp.
- The assistant introduces itself as a digital assistant of your company.
- It understands spoken answers, typed answers and WhatsApp voice messages.
- It learns the job titles and cities from the vacancy, so it hears them correctly.
- Audio is transcribed and only the text reaches the language model. Ringtime does no emotion inference and no voice biometric identification.
- The answers and the result land in the Ringtime web app and in your applicant tracking system (ATS).
To screen candidates through a real conversation, see how Ringtime's conversational screening works.
