AI for HR and recruiting
AI answers candidates and staff on WhatsApp and Telegram: it runs the first interview, books the next step, and explains how to request leave or a certificate. Your people still decide who gets hired and whose leave is approved
WhatsApp · employee
10:05Hi! I’d like to take leave from July 14 to 27. How many days do I have left, and what do I need to file?
HR agent
Reads and acts
Reading the question
- leave July 14–27
- days left
- how to file
HR system · Leave balance
Leave balance
Policies · How to file
How to file
Access and rules
Reply to the employee
10:05Hi Emma! You have 18 days left, enough for July 14–27. I’ve drafted the request and sent it to Daniel for approval. His answer will come to this chat.
Bitrix24 · Request for the manager
- Request:
- leave, July 14–27
- Approver:
- Daniel, store manager
Works with what HR already uses
-
Several languages
Handles mixed speech and one-word replies
-
People decide
A person signs off hiring, rejections and leave
-
Tested before launch
On your own candidate chats
What AI can take off HR’s plate
Start where a recruiter or HR officer writes the same thing every day: first interviews, bookings, questions about leave and certificates. This is usually where AI implementation in HR begins too
How the agent reasoned
AI already at work in HR
Both projects were built for Magnum, the largest retail chain in Kazakhstan. One agent takes candidates through to a trial day in a store, the other answers staff from internal documents
Scenario transcript
Application
- Under the hood: New application. HR system, Vacancy «Cook», Vacancy: Cook, WhatsApp: +7 747 ••• 19 03. Applied for “Cook”
- Agent: Good afternoon! I’m a recruiter at Magnum. You applied for the cook vacancy. Are you still looking for work?
Interview
- Customer: Yes
- Agent: Thank you! May I ask how old you are?
- Customer: 23
- Under the hood: Vacancy requirements. Age 23: fits. Age fits
- Agent: Where would you like to work? Tell me the district or address.
Nearby stores
- Customer: Alatau district
- Under the hood: Stores with the vacancy. Store map, Alatau district · Cook, Store 1: 1.2 km, Store 2: 2.8 km. Two stores nearby
- Agent: There are two stores nearby with an open cook vacancy: 1.2 km and 2.8 km away. Shall I book you an intro day at the nearest one?
Booking
- Customer: Yes, the nearest one
- Under the hood: Candidate’s answer. English, agrees, nearest store. Agrees to the nearest
- Under the hood: Intro day booking. HR system, Store: 1.2 km, Card: to the recruiter. Booking created
- System: Booking created · the recruiter got the candidate card
HR agent · WhatsApp · MagnumMagnum: store hiring over WhatsApp
An application in the HR system turns into a WhatsApp conversation. The agent runs the interview, picks the nearest store, explains the terms and books a trial day. The recruiter gets a candidate who is already screened.
- 2–3 nearest stores with an open role offered to each candidate
- 100 questions in Russian and Kazakh in the reference set
- Books the trial day right in the chat
- Screen recording from the live botStaff assistant · Telegram · Magnum
Magnum: policy answers for staff
An employee asks in their own words: how to request leave, who takes the sick note, where to get an employment certificate. The assistant finds the answer in approved documents and shows which document it came from.
- 100 real employee questions in Russian and Kazakh, run against every update
- When the documents have no answer, the assistant says so
- Reads tables and diagrams pasted into PDFs as images
Works on top of your HR systems
No need to replace your HR or payroll system. To start, a list of vacancies, locations and policies is enough. Writing into the HR system comes once the pilot has proved useful
- WhatsApp
Telegram- Careers page chat
HR agent
Understands the question, finds the answer in your data, and acts by HR’s rules
OpenAI
Anthropic
1C, payroll staff and leave reads - Bitrix24 requests and tasks reads writes
Google Sheets vacancies and shift rotas reads 
Notion, Drive policies and guides reads 
Your HR system applications and statuses reads writes
HR decisions stay with people
Before launch we agree with HR what the agent does on its own, what it only drafts for sign-off, and what goes straight to a person
Does on its own
Routine questions and bookings
- Explains schedules, terms and documents
- Runs the first interview with the recruiter’s questions
- Books interviews and sends reminders
- Answers staff from policies, citing the section
Drafts for sign-off
The draft is ready, a person sends it
- A rejection for a hard requirement
- Changes in 1C and HR records
- A message to a whole store or department
Goes straight to a person
With the chat and the facts collected
- Hiring, rejection and transfer decisions
- Pay, payments and disciplinary matters
- Health, conflicts, complaints about a manager
- Any request to talk to a person
We agree these rules with the HR director before launch. Every step the agent takes is logged, so you can always see why a candidate was offered that location or time
How we test an HR agent before launch
We agree the quality threshold with HR up front and test on real chats with your candidates and staff
- A set built from your chats
- 100–200 real questions from candidates and staff, with answers a recruiter approved. The set includes one-word replies like “ok” and “too far”, typos, Kazakh and mixed speech.
- A different model grades
- A separate judge model, not the one that answered, scores the answers. We check its scores against your staff’s own grades.
- Every miss gets reviewed
- An overall score can hide individual errors, so every answer below the threshold is reviewed on its own and fixed before launch.
- Acceptance on new questions
- We sign off on questions the system never saw during tuning, so the result can’t be fitted to familiar ones.
- WA Do you have night shifts? 0.93
- TG How do I request leave? linked the policy section 0.91
- WEB Do you hire without experience? 0.89
- WA How much is the pay during probation? judge: quoted a figure the vacancy does not have 0.66
- TG Who do I give my sick note to? 0.90
- WA Жұмыс кестесі қандай? 0.86
- TG I want to transfer to another branch handed to HR right away 0.94
- WEB What documents do I need to sign on? 0.88
- WA cant come tomorrow can we move it moved the interview in the calendar 0.85
- TG When is the advance paid? judge: the answer is not in the policies 0.71
Illustration of the mechanics with sample data. The real set is built from your requests during the audit.
How we launch AI in HR
Five stages. After each one you decide whether to go on. Each stage is paid separately
Pick one process
2–3 working daysWe find where the team loses hours and pick one process with an owner on your side and a result that can be counted.
A short memo: the process, the type of solution, what success means, and what stays out of the first stage.
Process, metric and owner are agreed.
Timelines start once access is granted: without chat exports and a test API the audit cannot start. WhatsApp number verification with Meta takes from a few days to three weeks, so we start it on day one. Projects with several processes or heavy integrations with 1C and custom systems take 3–4 months, but even there a working version ships in the first month.
Questions before an HR AI project
No. The agent collects the candidate’s answers, marks which requirements are not met, and writes a short summary. A person decides on hiring, rejection, transfer or discipline. The agent can draft a rejection for a hard requirement, but a recruiter sends it.
Before launch the recruiter writes down the must-haves for the role: schedule, age, documents, area. The agent asks one question at a time, checks the answers against those rules, and gives the recruiter a card: what fits, what to verify, where the risk is. Nobody has to read the whole chat.
From your HR system, a job board, or a form on your site. As soon as an application arrives, the agent messages the candidate on WhatsApp. In the Magnum project applications come from Magnum’s internal HR system, and the outcome of the conversation is written back there.
Yes. We choose a reliable integration path and discuss message templates, limits, consent, conversation storage, and when the dialog should move to a person. Employees’ private chats stay separate from the work system.
Yes. It replies in the candidate’s language and copes with mixed speech, typos and one-word answers like “ok” or “too far”. We test quality for each language separately: for Magnum the reference set had 100 questions in Russian and Kazakh.
The agent sees only the fields it needs to answer: leave balance, schedule, request status. It shares personal details only with the employee, identified by phone number. By default it does not discuss pay, sick leave or disciplinary matters and passes them to HR. Access is split by role, and every request stays in the log.
We separate access by role, limit approved sources, and log what the AI does. Financial, legal, HR, and other sensitive decisions require a human handoff. Disputed answers are reviewed against real conversations and added to the test set.
A candidate agent starts at $2,000, with a working version in 2–4 weeks. An employee assistant that searches your HR policies starts at $3,000 and takes 3–5 weeks. Support after launch costs $200 to $1,000 a month. We quote the first stage after a short review of the process.
A process owner who makes decisions, 30–100 real requests, conversations, or documents, a list of systems with access to their test APIs, handoff rules, and examples of good team answers. The knowledge base does not have to be perfect: we build the test set together.
Reading on AI in HR
- AI for HR What to give the agent, what to keep with recruiters, and how to run a 30-day pilot
- How AI saves recruiter time Where a recruiter’s day goes and which numbers to watch after launch
- Internal ChatGPT for a company How to answer staff from documents while respecting access rights
- What are AI evals and why projects need them How a test set catches agent mistakes before candidates see them
Let’s talk
Tell me who you hire, how many applications come in each month, and where your vacancies and policies live. I’ll suggest a first workflow you can test within a month.
I read every request myself.