AI for HR in a small company most often handles one specific task: the initial conversation with a candidate. A bot on Viber, Telegram or the website responds to applications within seconds at any time of day, explains the working hours and pay as per the job description, collects contact details and sends a completed candidate profile to the recruiter. The interview and the decision to hire remain the responsibility of a human.
Recruitment is currently the main area where HR teams are implementing artificial intelligence. According to the SHRM State of AI in HR 2026 study (a survey of 1,908 HR professionals, report dated 31 March 2026), 39 per cent of organisations use AI in HR functions, and 27 per cent in recruitment – which is more than for any other HR task. Below, we’ll look at how this plays out in a company where there is just one recruiter and applications are coming in thick and fast.
Contents
- Monday, 8.50 am: forty messages from candidates and not a single reply
- Between the interview and the first phone call, the candidate goes missing
- What AI for HR handles during the first interview
- The conversation on Viber is going on late into the evening
- The bot doesn’t decide who to choose
- The bot and the recruiter are each responsible for their own stage
- What information about a vacancy should you gather before launching it?
- The first week with the AI recruiter involves reviewing candidates’ conversations
- Frequently asked questions
Section: Instructions and cases
Monday, 8.50 am: forty messages from candidates and not a single reply
Imagine a warehouse complex just outside Kyiv. They need to recruit 12 order pickers before the season starts; the vacancy is advertised on Work.ua and on a local Telegram channel, with a Viber number at the bottom of the advert. Oksana is in charge of recruitment, whilst also managing HR paperwork and shift rosters.
On Friday evening, the advert was reposted in the local neighbourhood chat group. Over the weekend, I received forty messages on Viber. Most of them were the same: “What’s the salary?”, “Is transport provided?”, “Can I apply without experience?”, “Is it a formal job?”. A few people sent photos of their CVs, whilst others asked straight away if they could start tomorrow.
On Monday at 8.50 am, Oksana opens her messaging app and starts replying from the top down. By lunchtime, she will have got as far as the twentieth conversation. Meanwhile, some of those who messaged on Saturday have already received a reply from the neighbouring warehouse.
Between the interview and the first phone call, the candidate goes missing
Candidates are used to hearing nothing from employers, and this is where AI for HR makes its first noticeable difference. According to a survey Work.ua (November 2022, 9,142 candidates), 76 per cent of jobseekers had experienced a situation where a recruiter had not contacted them after an interview. Only 24 per cent always received a reply.
A more interesting explanation from employers: Work.ua surveyed a further 1,523 company representatives. 30 per cent cited the high volume of applicants as the reason, 22 per cent admitted frankly that they forget to reply, and a further 20 per cent blamed slow decision-making within the company. The first two reasons boil down to the same thing: people simply cannot keep up with the flood of messages.
At the initial response stage, this issue is more acute than after the interview. Someone looking for a job in a warehouse or in deliveries applies to several places at once. Whoever replies first and sets out clear terms is the one who gets the call.
What AI for HR handles during the first interview
The initial contact with a candidate consists of repetitive questions, the answers to which are already provided in the job description. It is precisely this part that is handled by the bot. Based on the job description, the Marichka AI manager responds within 10–30 seconds at any time of day or night, drawing its answers solely from the knowledge base you have provided. It is configured for job vacancies in the same way as for a product catalogue: instead of prices and stock levels, the database contains the terms and conditions of employment.
The bot is integrated where candidates are already messaging: on Viber, Telegram, Instagram, Facebook, WhatsApp or via a widget on the website’s job vacancies page. Simply replace the phone number in the job advert with a link to the bot on Viber or Telegram, and the candidate can start a conversation with just one tap.
In practice, the bot handles a number of things. It explains the shift rota, the pay rate, the address and how to get there. It answers questions about experience, age, documents and formal employment procedures. It collects the candidate’s name and telephone number and asks when it would be convenient for them to speak. If a candidate asks something that isn’t in the database, or clearly wants to speak to a human, the bot forwards the conversation to the recruiter along with the entire chat history.
The main difference from a button-based questionnaire bot is that this system understands free-form text. A candidate writes, “How much is the take-home pay per month, and do you pay for overtime?”, and receives an answer to both questions, without a “select option 1–5” menu. We explain how such systems work and what their limitations are in this article on AI agents and one-click bots.
The conversation on Viber is going on late into the evening
Here’s what that Saturday review looks like when the advert contains a link to a Viber bot instead of a phone number.
The applicant writes at 10.15 pm: “Good evening. Are you still looking for order pickers? I have no experience.” The reply comes within a few seconds: yes, the vacancy is open, no experience required, and there’s a mentor for the first week. The bot immediately gives a brief summary of the rota and pay rate from the job description.
Next, the candidate asks about transport. The bot replies, stating where the company bus departs from and at what time. The next question is: “What if I’m a student – can I only work night shifts?” Here’s a rule from the knowledge base: the recruiter decides on individual working hours. The bot says exactly that, asks for a name and phone number, and suggests a time for a call.
On Monday at 8.50 am, instead of forty half-baked conversations, Oksana sees a few messages with context: who, where they’re from, what they’ve already asked, and when it’s convenient to speak. The standard questions had already been sorted out at the weekend. All she has to do is ring those who are ready to come in and start with the earliest ones.
If your weekends currently look like the first option, it’s worth applying for one vacancy before the seasonal recruitment drive begins.
The bot doesn’t decide who to choose
The biggest temptation with an AI recruiter is to let it filter out candidates who “don’t fit the mould”. It is best not to do this. The EU Regulation on Artificial Intelligence (2024/1689, Annex III, point 4(a)) classifies systems for recruiting staff as high-risk, particularly those that analyse and filter applications and assess candidates. Ukraine is not part of the EU, but the logic behind this provision is clear: automatically rejecting a person without explanation is precisely the kind of action that most easily amounts to discrimination.
That’s why there’s a simple rule for the HR bot. It provides information and gathers data; it does not make judgements. If a candidate does not meet a formal requirement for the vacancy (for example, a medical certificate is required but they do not have one), the bot honestly states the requirement and still passes the contact details on to the recruiter. The decision is made by a human.
The second rule concerns data. Before a candidate hangs up, they must understand what they are writing to the automated assistant and consent to the processing of their personal data. Access to correspondence with candidates in Marichka AI is role-based, so only those involved in the recruitment process can view the conversations.
The bot and the recruiter are each responsible for their own stage
It is easier to agree on the division of tasks in advance than to sort out every contentious conversation. An HR chatbot handles all repetitive tasks for which there is a ready-made answer, whilst anything requiring human judgement is left to the recruiter.
| Stage | Bot | Recruiter |
|---|---|---|
| First reply to a review | Available 24/7 | Sees the dialogue in the morning |
| Enquiries regarding terms and conditions, timetable and address | Matches the job description | Updates the knowledge base when conditions change |
| Collecting contact details | Name, telephone number, preferred time for a call | Checks the card before making a call |
| Non-standard enquiry | Transfers the person along with all correspondence | Makes decisions and takes personal responsibility |
| The interview and the decision to hire | Not taking part | Conducts and adopts |
What information about a vacancy should you gather before launching it?
The bot knows exactly what is in its database. If the job description makes no mention of a lift, it won’t come up with a route and will pass the query on to a human. This is safe, but every such handover reduces the overall benefit. That’s why the database should be compiled once, but comprehensively. The success of AI for HR depends more on this preparation than on the bot’s settings. With a half-page job description, the bot will have to contact the recruiter for almost every question.
Minimum requirements for one vacancy:
- job description, shift rota, pay rate and how overtime is calculated;
- address, how to get there, whether there is a company transport service;
- eligibility criteria and documents required for application;
- answers to 20–30 questions that candidates have already asked via the messaging app;
- Rules for handing over the conversation: which topics the bot does not discuss and to whom it hands over the conversation.
The best source for the last two points is your own correspondence from the previous season. Go through it and note down the questions that came up most often. Marichka AI can import PDF or Excel files without any coding, and it can read job descriptions from websites itself.
The first week with the AI recruiter involves reviewing candidates’ conversations
For the first seven days, read through the conversations every day, even if only selectively. It is this morning review of the conversations that determines whether the bot will respond better after a month than it did on the first day. A general impression of “does it work or not?” isn’t much use here. Look at the questions where the bot repeatedly hands the conversation back to a human. Check separately whether candidates get stuck at any particular stage, for example, when the bot asks for a phone number. And cross-check the answers about pay with what Oksana says in the interview, and the rota against the actual shift schedule.
Every repeat query asking the same question is a ready-made entry for the knowledge base. Add a reply, and next time the bot will close it automatically. Sort out any discrepancies in the terms straight away: a candidate who was quoted one salary in the chat but another at the interview won’t come back a second time.
For requests outside of HR, such as orders placed via a messaging app, the trigger logic is the same. For other tasks, see the article on… What other tasks does a chatbot handle in small businesses?.
AI for HR in a small company makes sense where candidates apply outside working hours and there is only one recruiter. Start with a single vacancy involving mass recruitment and one channel that generates the most responses, and after a week’s trial, compare how many candidates made it to the phone interview with the figures from a typical weekend.
Ready to connect Marichka to your channel?
Marichka AI will help you automate communication, reply to clients 24/7 and increase sales.
Frequently asked questions
Will AI replace recruiters in small companies?
No. The AI handles the initial contact: it answers questions about the vacancy, working hours and pay, and collects the candidate’s name, telephone number and a convenient time for a call. The interview, candidate assessment and hiring decision are carried out by a human. The recruiter has fewer repetitive conversations and more time to focus on those who are already ready to attend.
Is it legal for a bot to communicate with a candidate?
It is possible to communicate via a bot provided that the candidate understands they are writing to an automated assistant and consents to the processing of their personal data before providing their telephone number. Be cautious when assessing people: under the EU’s Artificial Intelligence Regulation, systems that filter applications and assess candidates are classified as high-risk. Therefore, leave the decision on the candidate to the recruiter.
How long does it take to set up an HR chatbot?
It takes just a few minutes to set up the channels and the bot. The most time-consuming part is compiling the knowledge base: job descriptions, terms and conditions, addresses, and answers to candidates’ frequently asked questions. If these documents are already in PDF or Excel format, they can be imported into Marichka AI without any coding. You can start testing with a single job vacancy and a single channel during the 7-day trial.
