Alexandr Chibilyaev on AI agents that screen resumes, schedule interviews, follow up with candidates, and manage onboarding โ powered by integrations with HH.ru, Huntflow, SuperJob, and a full HR connector stack.
Every recruiter knows the feeling: 200 unread resumes, 47 candidates at different interview stages, 14 follow-ups that should have gone out yesterday, and a hiring manager asking "what's the status on the senior developer role?" for the third time this week.
Recruiting doesn't lack tools. It lacks connected tools that talk to each other and make decisions without human intervention for every micro-step. An ATS stores candidates. A job board delivers applications. A calendar holds availability. An email client sends follow-ups. The recruiter is the integration layer โ and that's the problem.
AI agents don't replace recruiters. They replace the integration layer. They connect HH.ru to Huntflow to SuperJob to Telegram to Google Calendar. They read resumes, score candidates, schedule interviews, and follow up โ not as demos, but as production infrastructure that handles 80% of the repetitive work so recruiters can focus on the 20% that requires human judgment: assessing culture fit, negotiating offers, and building relationships with top candidates.
AACFlow connects the entire HR ecosystem through our knowledge base sync architecture. The agents don't "call the HH.ru API." They understand the talent pipeline โ every candidate, every resume, every interaction โ because the entire hiring workflow lives in their working memory.
Each connector normalizes candidate data into a unified model. A resume from HH.ru, a response from SuperJob, a candidate profile from Huntflow โ all mapped to the same fields: name, contact, skills, experience, salary expectations, current stage. The agent searches one knowledge base, not four separate APIs.
This is where recruiting time goes to die. A job is posted. 150 resumes arrive. Someone needs to open each one, read it, compare it to the job description, and decide: yes, maybe, no. For every 150 applications, maybe 15 are worth a phone screen. The other 135 are a tax on attention โ but you still have to read them to find the 15.
The Resume Screening Agent automates the first pass:
Ingest โ webhook fires when a new application arrives from HH.ru or SuperJob. Resume text, cover letter, and test results (if any) are pulled into the knowledge base.
Parse โ the agent extracts structured data from unstructured resume text: skills (technologies, languages, frameworks), years of experience, education, previous companies and roles, salary expectations, notice period, location, work authorization.
Score โ the agent compares the candidate profile against the job requirements using a configurable rubric:
Hard requirements (must have): specific technologies, years of experience, language proficiency. Missing a hard requirement โ auto-reject with a polite notification.
Soft preferences (nice to have): industry experience, specific tools, optional skills. Contribute to score but don't disqualify.
Negative signals: unexplained employment gaps (flagged, not auto-rejected โ the candidate might have valid reasons), frequent job changes, salary expectations >2x above budget.
Rank โ candidates are sorted into three tiers:
Tier 1: high match, auto-advance to phone screen stage in Huntflow
Tier 2: moderate match, flag for recruiter review
Tier 3: low match, polite rejection with option to join talent pool for future roles
Justify โ for every scoring decision, the agent writes a 3-line rationale: "Strong match on Python (6 years) and PostgreSQL (5 years). No experience with Kubernetes (required). Previous role at Yandex signals strong engineering culture fit."
The recruiter sees a ranked, annotated shortlist โ not a pile of PDFs. The agent handled the reading. The recruiter handles the deciding.
The second-biggest time sink in recruiting: communication. A candidate applies. You need to: confirm receipt, schedule a call, send the meeting link, follow up after the interview, send the test assignment, remind them about the deadline, thank them for completing the test, schedule the technical interview, send the interviewer briefing... and that's just for one candidate at one role.
The Candidate Communication Agent handles every touchpoint:
Application confirmation โ instant automated response when an application arrives, personalized with the role title and expected timeline
Screening invitation โ for Tier 1 candidates, the agent sends a message (HH.ru messenger, Telegram, or email) with available time slots pulled from the recruiter's calendar
Follow-up sequences โ if a candidate doesn't respond within 48 hours, the agent sends a polite follow-up. After 7 days with no response, the agent marks the candidate as "no response" in Huntflow and stops the sequence.
Post-interview updates โ after each interview stage, the agent sends a status update: "Great conversation with Dmitry! We'll be moving you to the technical interview stage. Expect an invitation within 24 hours." Or: "Thank you for taking the time. The team has decided to move forward with other candidates for this role. We'd love to keep you in our talent pool for future openings."
Offer communication โ when an offer is approved in Huntflow, the agent generates the offer letter (with specific compensation, start date, and benefits pulled from the vacancy record), sends it via the candidate's preferred channel, and tracks acceptance deadline
Rejection with dignity โ rejected candidates receive a thoughtful, personalized message that references specific strengths from their profile, not a template. The agent reads the resume and writes: "Your experience scaling systems at [Company] is impressive. While we went with a candidate who had deeper Go experience for this particular role, we'd love to stay in touch for future backend positions."
Every message is in the company's voice, defined in the agent's system prompt. The agent doesn't sound like a bot because it isn't generating template messages โ it's reasoning about the specific candidate and writing to them.
The scheduling coordination tax. Recruiter emails candidate: "When are you free?" Candidate replies: "Tuesday after 2 PM or Thursday morning." Recruiter checks calendar. Tuesday is booked. Thursday the hiring manager is out. Recruiter emails: "How about Wednesday at 3?" Candidate replies: "Can we do 3:30?" And so on. Five emails to settle one time slot. For 20 interviews per week, that's 100 emails of pure coordination.
The Interview Scheduling Agent eliminates the back-and-forth:
Access calendars โ the agent reads the interviewer's calendar (Yandex.Calendar or Google Calendar via connector) and the candidate's availability (collected during application or provided via a scheduling link)
Find slots โ the agent identifies overlapping windows, respecting:
Buffer time before/after interviews (15 minutes)
Interviewer working hours
Time zone differences
Room availability (if in-office interviews)
Propose โ the agent sends the candidate 3 specific, ready-to-book time slots via their preferred channel
Confirm โ when the candidate selects a slot, the agent:
Creates the calendar event for both parties
Books the room (via the office management connector, if integrated)
Updates the candidate's stage in Huntflow to "Interview Scheduled"
Sends a confirmation with the meeting link, interviewer name and title, and a 2-line description of what to expect
Remind โ 24 hours and 1 hour before the interview, the agent sends a reminder to both parties. For the interviewer, the reminder includes a candidate briefing: resume highlights, previous interview notes, specific points to probe.
Handle reschedules โ if the candidate needs to reschedule, the agent handles it. No human in the loop. The agent picks new slots, sends them, processes the selection, updates the calendar.
Twenty interviews per week, zero scheduling emails. The recruiter checks their dashboard and sees: 18 interviews confirmed, 2 awaiting candidate selection. No coordination work.
A candidate accepts the offer. Now the real paperwork starts: employment contract, NDAs, tax forms, bank details, passport copy, education certificates, medical records. Each document needs to be requested, received, verified, filed. Miss one document, and payroll can't process the new hire's first salary. That's not a small mistake โ it's a relationship-damaging failure in the first month.
The Onboarding Document Agent manages the entire document collection lifecycle:
Generate checklist โ based on the role, location, and employment type, the agent generates a personalized onboarding document list: which documents are required, which are optional, which need wet signatures, which can be digital
Request โ the agent sends the new hire a structured request with clear instructions, deadlines, and upload links (via Telegram bot or secure document portal)
Track โ the agent monitors completion status. Documents not received by their deadline trigger a polite reminder. Documents flagged as "missing information" trigger a specific request for the missing detail.
Verify โ the agent performs first-pass document verification:
Passport: checks that the name matches the offer, the passport isn't expired, the photo is clear
Tax documents: verifies INN/SOANO format for the relevant tax jurisdiction
Employment record: checks for completeness (all pages present, signatures where expected)
Education certificates: verifies the institution name and degree against the candidate's stated qualifications
Escalate โ documents that fail verification are flagged for HR review with specific, annotated reasons ("passport photo is blurry โ request rescan" vs. "education certificate from unverifiable institution โ require additional documentation")
File โ verified documents are stored in the knowledge base with appropriate access controls, linked to the employee record, and indexed for future reference
The agent doesn't make final approval decisions on sensitive verification. It does the first pass โ the reading, the checking, the flagging โ and surfaces only the exceptions. HR reviews the flagged items, approves the rest, and the new hire is ready for payroll on day one.
Humans handle: assessing culture fit, making hiring decisions, negotiating offers, having sensitive conversations, evaluating ambiguous qualifications, deciding when to override the agent's scoring, building relationships with top candidates.
The agent recommends. The human decides. This isn't a limitation. It's the design. Hiring is one of the highest-stakes decisions a company makes. Automation should handle the volume so humans can focus on the judgment.
Every agent action that touches a candidate directly can be configured for:
Automatic mode: the agent acts immediately (application confirmations, document reminders)
Review mode: the agent drafts and queues for human approval before sending (rejection messages, offer letters)
Advisory mode: the agent only provides recommendations to the recruiter (candidate scoring, interview briefing)
You choose the mode per workflow, per stage. Control stays with the people who are accountable for hiring outcomes.
92% of incoming applications screened within 5 minutes โ previously, initial screening took 2-3 days due to recruiter backlog
74% reduction in scheduling coordination time โ from an average of 7 messages to book an interview down to 2
Zero missed follow-ups โ the communication agent never forgets a candidate, never lets a thread go cold
Onboarding document collection complete 4 days earlier on average โ from 12 days to 8 days from offer acceptance to full document readiness
Recruiters report 40% more time spent on candidate engagement and relationship-building โ the work that actually improves hire quality
One tech company with 30+ open roles told us: "Before AACFlow, our two recruiters were drowning in screening and scheduling. They were spending maybe 20% of their time actually talking to qualified candidates. Now screening is automated and scheduling is one click. They spend 60% of their time on high-value interactions. Our offer acceptance rate went up because candidates felt more engaged throughout the process."
Connect your hiring platforms โ HH.ru, SuperJob, and Huntflow from the connector catalog
Connect your communication channels โ Telegram for candidate messaging, email for formal communications
Connect calendars โ Yandex.Calendar or Google Calendar for scheduling
Deploy agents โ start with resume screening (biggest time impact, lowest risk), then add communication and scheduling
Set your human-in-the-loop boundaries โ which actions auto-execute, which require review, which are advisory only
Recruiting is a human process augmented by technology โ not a technology process that happens to involve humans. AACFlow HR agents handle the operational volume so recruiters can invest their time where it creates the most value: building the team.
