An AI assistant is writing to you about a job: what it is and who reads the chat

Updated

Crutor is a service recruiters use to screen applications in Telegram: an AI assistant asks you the job's questions on behalf of a real hiring team, and that team reads the whole chat. The AI doesn't hire or reject anyone: it collects your answers and passes on what it can't answer, and the people on the team decide.

What is Crutor, and why is it writing to me?

A recruiter or a company creates a job in Crutor, writes the questions they want answered, and shares a link to the bot @Crutorbot in a job post or a message. You get messages from Crutor in one of two ways:

  • You opened a job link (it starts with t.me/Crutorbot) and pressed Start. A Telegram bot can't write to anyone first, so this chat exists because you started it.
  • You wrote to a recruiter's own account, and that recruiter connected Crutor through Telegram Business. Then the assistant answers in that chat under the recruiter's name: “Hi! I'm …'s assistant. Is this about the “…” job?”

The first message names the job and, unless the recruiter hides it, the company. It also says “The team hiring for this job will read your messages” and gives a link to the privacy policy.

The first message names the job, the company and who reads the chat; asked whether it's a bot, the assistant says so plainly.

How to tell it's real. The bot's username is exactly @Crutorbot. The hiring team never asks for money, card details or codes from Telegram or SMS. If someone in the chat does, don't pay or send anything: tell the bot, and it will warn the team.

Is it a bot or a person?

Most messages are written by an AI. If you ask, it says plainly that it is the hiring team's AI assistant and that the team reads the conversation: it is instructed never to claim to be human.

People from the team can write to you in the same chat at any time. Their messages also come from the bot, worded as they wrote them; in a Telegram Business chat they come from the recruiter's own account.

Who reads this chat?

  • The hiring team for this job: the recruiter and the members of the job's Telegram group. Everything you send (text, files, voice messages) is copied into a topic of that group, one topic per candidate, and is in the team's web app. They see your answers, your documents, a summary and the AI's score.
  • You see your own answers and can correct them. You never see the score, the team's notes or other candidates.
  • Crutor's administrators, only to keep the service running, to help users and to look into abuse or a complaint. Every time an administrator opens a conversation, it is logged.
  • The AI providers that write the replies process the text. Under Crutor's terms with them, they don't train their models on it.

Nobody else. Crutor doesn't sell data or use it for advertising.

What does the assistant do, and what doesn't it do?

It does It doesn't
Ask the job's questions one at a time, in its own words Decide whether you're hired or rejected
Answer questions about the job, only from the job's own text Tell you a score or hint at a decision
Pass a question it can't answer to the team Invent dates, pay or promises the job text doesn't give
Take your CV, files and links; voice messages are transcribed Ask for money, card details or codes
Show you a summary of your answers to check before anything goes to the team Send links to other sites
Offer interview times when the team invites you, and book the one you pick Keep asking once the team has paused it

Each reply is checked before it is sent. A reply that states a hiring decision, asks for codes or money or contains a foreign link is not sent: you get a neutral reply instead, and the team is asked to answer you itself.

How is the decision made?

  1. You confirm the summary. When all the required answers are in, the bot shows them with “✅ All correct, send” and “✏️ Change something”. The application goes to the team only after you confirm.
  2. The AI scores it for the team. It rates the fit from 1 to 10 against the job's questions and lists strengths and concerns. Only the team sees this.
  3. Deal-breakers mark the card, they don't reject. A recruiter can mark a question as a deal-breaker, for example “no smartphone” for a courier. If your answer matches, the team's card shows it under “Deal-breakers” and the score is low. Nobody is rejected automatically: the application's stage changes only when someone on the team presses a button.
  4. A person decides. The team moves the application to the shortlist, an interview, hired or rejected.
  5. A rejection message is the team's choice. When someone rejects an application, the bot asks them whether to message the candidate. A rejection reaches you only if they choose to send one; the AI never sends one on its own.

You can still change an answer after sending; the score is then recalculated.

How do I reach a person?

Write it in the chat in your own words: “I'd like to talk to a person”, or simply your question. When you ask for a person, or ask something the job text doesn't answer, the assistant passes your words to the team's topic and tells you the team will reply in this chat. The reply comes here, in the same chat.

How fast the team answers is up to the team. For problems with Crutor itself rather than with the job, write to support@crutor.app: a person reads it.

What does “the team will reply to you personally” mean?

The team can pause the AI in your chat, for example to talk to you themselves. While it is paused, the AI doesn't reply, and your messages wait for the team.

  1. The first time you write during a pause, you get one notice: “Thanks, your message is received. The team will reply to you personally — it may take a little while.”
  2. It comes once per pause, not after every message.
  3. The team answers in the same chat. When they turn the AI back on, it picks up the messages that waited.
The team has paused the AI: your message gets a one-time notice, and the next reply comes from a person on the team.

The AI also pauses by itself if one chat gets too many replies in a day, with the same notice. In a Telegram Business chat, the AI steps aside as soon as the recruiter writes to you themselves; there is no notice there, because you see them writing.

Which language is the chat in?

Crutor's own messages are in English, Russian or Ukrainian. The chat starts in the language of the job text and then follows the language you write in. Other languages aren't supported: a message in Latin letters gets replies in English, one in Cyrillic in Russian, or in Ukrainian if it has Ukrainian letters.

Your answers reach the team as you wrote them.

How do I stop?

  • Withdraw: send /withdraw and confirm. The team sees that you withdrew; if you change your mind, the button under the bot's “withdrawn” message takes you back, answers and all. /applications lists your applications.
  • Just stop answering: if you leave an application unfinished, the bot reminds you what is left after about an hour and once more a day later, never at night in your time zone, and then nothing more.
  • Block the bot in Telegram: it can't write to you at all, and neither can the team through it.