Hiring with an AI recruiter: a job opening is an empty seat

The head of logistics has been trying to hire a dispatcher for three months. The job is posted on two sites, resumes land in a shared inbox, and a few arrive in someone’s personal messenger. HR forwards the promising ones, the manager replies a day later, and two strong candidates have already accepted offers elsewhere. When the right person is finally found, round two begins: the employment paperwork, payroll setup, system access. Each step involves a different person and a different tool, and something always falls through the cracks.

PRANUS handles hiring differently. A job opening isn’t a post on a careers page. It’s an empty seat on your org chart. From that seat, an AI recruiter starts sourcing, holds the first conversation with candidates, and prepares a brief for the hiring manager. When someone is hired, the seat on the chart becomes an employee, with paperwork, salary, and access already in place. And every decision that matters stays with people.

A job opening starts on the org chart

In TEAMUS, the org chart is a living canvas that shows empty seats as well as people. An empty seat is the opening. The manager opens it and clicks “Find someone.” The department, the manager, and the job description fill in automatically, because the system already knows where the person is needed and what they’ll do.

The salary range must be approved by someone other than the person who proposed it, a simple rule against bias. Only approvers and the hiring manager can see it, not everyone involved in screening.

The profile: who you actually need

Next, the AI has a short “who do we need” conversation with the manager. It isn’t a forty-field form. It’s a discussion about what results you expect in three months, what’s non-negotiable, what’s preferred, what’s a bonus, and what’s a dealbreaker. Location and work format, experience, languages, pay expectations, and first-conversation questions get added along the way.

The result is a three-tier set of criteria:

must have: we won’t hire without it;
nice to have: we’d like it;
bonus: a pleasant extra.

Each criterion comes with a note on how to verify it. Gender, age, and family status aren’t stored at all. They’re not part of the profile, and screening never uses them.

If your company has written down its values, the profile builds on them instead of reinventing them for every role.

Sourcing: your own pool first, then Telegram

Start with people you already know

The cheapest candidates are the ones you’ve met before. Sourcing starts with your talent pool and past applicants, especially near-misses for similar roles. One candidate record can be linked to several openings, and duplicates are detected automatically.

Public Telegram channels

Next comes scanning of public Telegram channels and groups where people post resumes or say they’re looking for work. The system picks channels to match the role and its criteria, and scores each channel before it starts reading. The manager sees a “channels for this opening” list and can remove one or add their own. We cover how we use Telegram without spamming anyone in Telegram lead generation without spam.

Candidates are sorted into three groups, good fit, maybe, not a fit, with an explanation for every criterion and a link to the original post or resume. The manager’s yes and no marks teach the scoring for that opening.

The first conversation: honest and to the point

Who the recruiter contacts first

Only people who posted their own resume or a “looking for work” message with contact details. They’re expecting to hear from employers. The recruiter doesn’t reach out cold to anyone else. The message is short, sent on behalf of the company, names the role, and mentions what stood out in the candidate’s post. It’s written in the candidate’s language.

The recruiter says it’s an AI

The first message never pretends to be a person. There’s no “Hi, I’m Jessica.” By the second message, the recruiter openly says it’s an AI assistant, explains why, and lets the candidate know their answers are stored. If the candidate asks “Are you a bot?” earlier, the answer “Yes, I’m an AI assistant” comes right away.

We know this may cost some replies. It’s a deliberate choice: honesty matters more than conversion. A candidate who’s misled in the very first message won’t become a great employee, and a company that starts relationships that way earns the reputation that goes with it.

What the recruiter will and won’t talk about

It asks first-conversation questions one at a time.
It answers questions about the company and the role only from the job card and the knowledge base. If it doesn’t know, it promises to check and asks a person.
On pay, it shares only what the manager marked as OK to tell candidates.
The candidate can ask for a human at any point.
If the candidate asks, their data is deleted right away.
If a candidate goes quiet, it follows up at most twice, about a day later and then after three or four days. After that, they move to “No response.”

The conversation ends with a brief for the manager: what’s confirmed, what the candidate didn’t address, any concerns, and verbatim quotes.

Screening on quotes, not a match score

Many recruiting tools show a neat number: “87% match.” We deliberately don’t. Once a percentage appears on screen, people stop reading and start comparing numbers, and the number hides what it’s based on.

In PRANUS, a criterion counts as met only with a verbatim quote from the candidate, from their resume or from the conversation. No quote means “not mentioned in the resume,” not “not a fit.” The manager sees the evidence rather than a score, and decides for themselves whether a given point matters.

Interview, work sample, personality

Then come the things that actually predict success: a structured interview with the same questions for every candidate, and a work sample tailored to the role. With consent, a candidate can also take a Big Five personality questionnaire. The result isn’t a filter. It’s a source of interview questions, like “You said you prefer clear instructions. What do you do when there aren’t any?”

Decisions stay with people

The AI recruiter never hires anyone and never makes an offer. It only moves candidates through the early stages:

Found → Contacted → Conversation with assistant → Passed screening.

Everything after that is done by a person: Manager interview → Work sample → Offer → Hired.

When a candidate passes screening, the manager gets a notification with the brief and three buttons: Interview, Talent pool, or Decline. Click Interview and the recruiter offers the candidate times from open slots on the manager’s calendar, then sends a reminder the day before.

The manager confirms rejections in a single daily batch instead of one at a time. That saves time and ensures the recruiter never turns someone down without a person knowing.

Hiring: the empty seat becomes an employee

Once the candidate accepts the offer, nothing has to be done by hand:

a “hire” event is created in TEAMUS;
DOCUS generates the employment documents under the employer’s country rules;
salary is set up in BUHUS accounting;
the new employee gets an invitation to the system;
they receive their department’s permissions in every service;
the empty seat on the org chart becomes an employee.

For how permissions flow from a seat on the chart, see An org chart that manages permissions.

Candidate data and privacy

Recruiting means personal data, so we store the minimum needed: name, contact details, where we found them, the conversation, and answers by criterion. People who weren’t a fit are kept for three months, and the talent pool only with consent. We don’t collect candidate video or voice. Every action the recruiter takes is logged: what it wrote, what it recorded, and where it moved the candidate.

Every opening has a spending cap for AI usage, so the recruiter can’t quietly spend more than you’ve allowed.

What changes for the hiring manager

In short, the hiring manager stops sorting resumes and starts making decisions. Here’s where their time goes:

Step	The usual way	With the AI recruiter
Define who you need	Write the job post or ask HR to	Answer questions in the “who do we need” conversation
Find candidates	Wait for applicants, browse job sites	Get three sorted groups with explanations
First conversation	Call everyone or hand it to HR	Read a brief with quotes
Schedule an interview	Back-and-forth about times	Click Interview, and the time gets arranged
Process the hire	Paperwork, salary, access, each separately	One “hire” event

What’s left are the parts that matter most: meeting the candidate, the work sample, and the decision. Exactly the parts that need a human.

How to get started

The recruiter is an AI employee configured in AGENTUS like any other agent, by describing it, without a developer. To get it running, you need three things:

an org chart with empty seats, which the agent builds during the PRANUS implementation;
a job description for the seat, which the AI drafts based on your structure and you edit;
your values and communication rules: what candidates can be told about pay, when the recruiter discloses it’s an AI, and what tone to use.

Want to see it on one of your open roles? Book a demo and we’ll open an empty seat on the chart and follow it all the way to the manager’s brief.