Guide: AI in business › AI in sales and business departments
AI in HR works best where it organises information: in job ads, answers to candidates’ questions, interview summaries and the analysis of employee feedback. It calls for the most caution where it assesses people. Under the AI Act, candidate selection and employee assessment are a high-risk area, with full obligations from 2 December 2027. Below we describe what works, what is already prohibited today, what will change, how much it costs and how AI affects the employer’s image.
In 60 seconds HR and AI at a glance
- Scope: artificial intelligence in recruitment, in communication with employees and in HR data analysis.
- Who will find it useful: HR departments that receive a lot of applications, run many recruitment processes at once or want to reply to candidates faster.
- Cost: general tools from a few dozen dollars per person per month, and the biggest costs are preparing the rules, training the team and quality control.
- A safe start: tasks that do not assess people, i.e. job ads, answering questions and organising notes.
- Hard limits: emotion recognition at work has been prohibited since 2025, and CV screening by AI is a high-risk system.
What AI in HR is and why the topic has come back right now
Artificial intelligence in HR covers all the tools that help in working with people, from a chat that answers candidates’ questions about the process, through summaries of job interviews, to systems that organise applications. It is not about one tool. It is about a set of tasks in which a program can take work off the recruiter’s hands. How HR fits into the broader picture of AI in a company, from legal to sales, is shown in our guide to artificial intelligence for business departments.
The topic has come back for two reasons. Firstly, candidates are using AI on a massive scale. According to a Pracuj.pl survey from May 2026, 48% of them used AI at some stage of recruitment. Secondly, the AI Act timetable has stabilised. The Digital Omnibus regulation, published on 24 July 2026, moved the full obligations for AI in recruitment to 2 December 2027. So there is plenty of time. But not unlimited time.
What AI is useful for in HR and recruitment
Zastosowania AI w HR (w ogłoszeniach często jako „AI HR”) najlepiej podzielić według ryzyka. Im mocniej narzędzie wpływa na decyzję o konkretnym człowieku, tym więcej obowiązków i tym większa odpowiedzialność pracodawcy, który nie może się potem tłumaczyć, że „tak wyszło z systemu”, bo dla kandydata, sądu i urzędu decyzję zawsze podejmuje firma, a nie program. Tabela pokazuje, jak to wygląda w praktyce.
| Application | Risk level | Notes |
|---|---|---|
| writing and editing job ads | low | check the language for exclusionary wording |
| answers to common candidate questions | low | the chat must disclose that it is a program (Article 50 of the AI Act, from 2 August 2026) |
| interview summaries and notes | medium | candidate consent to recording, data protection |
| analysis of anonymous employee surveys | medium | aggregate data only, no assessment of individuals |
| CV screening and ranking | high | high-risk area under the AI Act; the decision always stays with a person |
| employee evaluation, promotions, dismissals | high | high-risk area under the AI Act |
| emotion recognition at work | prohibited | banned since 2 February 2025 (Art. 5), with medical and safety exceptions |
Candidates use AI too: CV builders and what to do about them
An AI CV builder, a tool that writes a ready-made CV and cover letter on the basis of a few answers, has become something ordinary for candidates. Every recruiter sees the effect. Applications are correct and polished, and also as alike as two peas in a pod, so it is becoming increasingly difficult to tell a good candidate from a well-written document, especially when several hundred applications for one position come in within a week.
Recruiters accept this to varying degrees. In the same Pracuj.pl survey, 57% of recruiters agree to AI help in writing a CV and 67% in preparing for an interview, but only 3% accept the use of AI while answering during the interview. At the same time, only 7% of companies have rules on the use of AI and tell candidates about them. A recipe for misunderstandings.
Co z tym zrobić? Nie walcz z narzędziami. Zmień sposób oceny: mniej wagi do stylu CV, więcej do krótkiego zadania praktycznego i rozmowy o konkretnych sytuacjach, na przykład „opowiedz o ostatnim projekcie, który poszedł nie tak, i co wtedy zrobiłeś”. I napisz wprost w ogłoszeniu, na co pozwalasz. Kandydat, który wie, czego się spodziewać, wychodzi z lepszym wrażeniem, nawet jeśli odpadnie. Jak ustawić kryteria tak, żeby ocena służyła temu, czego firma naprawdę potrzebuje, opisujemy w tekście o selekcji pracowników zgodnej z celami firmy.
What the AI Act says about AI in recruitment
The AI Act (the EU regulation on artificial intelligence) is coming into force in stages, and four dates matter for HR. Status as of 24 September 2026, after the amendments made by the Digital Omnibus regulation (EU) 2026/1744:
- From 2 February 2025, already in force. A ban on emotion recognition in the workplace, except for medical and safety purposes (Article 5), and an obligation regarding AI literacy among the people who use it (Article 4). After the Digital Omnibus amendments, the employer is to “support the development” of these skills rather than guarantee a specific level.
- From 2 August 2026, already in force. Transparency: a candidate talking to a chat must know that it is a program (Article 50).
- From 2 December 2026. New prohibitions from the Digital Omnibus (concerning the generation of illegal intimate content, not recruitment), as well as the deadline for labelling content generated by systems that were on the market before August 2026.
- From 2 December 2027. Full obligations for high-risk systems listed in Annex III, including AI for recruitment, selection, candidate assessment, decisions on promotion and termination, and employee monitoring. An employer using such a system must, among other things, ensure human oversight, inform employees and keep records of its operation.
Jak Commwork®AI pilnuje zgodności AI w rekrutacji
Rekrutacja z AI niesie dwa ryzyka: przepisy i markę pracodawcy. Commwork®AI pokazuje całą drogę kandydata od ogłoszenia do końca okresu próbnego i łączy ją z programami, których HR już używa. Maszyna porządkuje, zespół Commplace® pomaga ustawić zasady, a o każdym kandydacie decyduje człowiek.
- Portale z ogłoszeniami 41%
- Polecenia pracowników 24%
- Strona kariery 18%
- social media 11%
- Agencje rekrutacyjne 6%
- 2.02.2025Zakaz rozpoznawania emocji w miejscu pracy (art. 5 AI Act)
- 2.08.2026Czat z kandydatem musi mówić, że jest programem (art. 50)
- Week 2Mapa procesu i lista narzędzi AI używanych w HR
- Tydzień 5Zasady AI dla kandydatów na stronie kariery
- Week 12Przegląd rezygnacji, ankieta i decyzja o dalszych krokach
- 2.12.2027Pełne obowiązki dla AI w rekrutacji i ocenie pracowników
- 01Mapa procesu i narzędzi
- 02Zasady i zgodność
- 03Droga kandydata w kokpicie
- 04Implementation
- 05Pomiar rotacji
- Łączy ogłoszenia, system rekrutacyjny i dane kadrowe w jeden widok drogi kandydata, od źródła zgłoszenia do końca okresu próbnego.
- Pokazuje nastroje w zespołach i sygnały odejść zbiorczo, bez oceniania pojedynczych osób przez maszynę.
- Porządkuje rekrutację pod unijne rozporządzenie o AI: człowiek decyduje, a system zapisuje, kto, kiedy i na jakich danych. Kalendarz obowiązków rozpisaliśmy w przewodniku The AI Act in Poland.
The view reproduced is modeled after the dashboards provided to customers. The names, numbers, and terms are examples and show the way the system works, not the results of a specific company.
Who AI in HR is for and when it pays off
The companies that gain most are those that run many similar recruitment processes: retail chains, manufacturing plants, logistics companies, service centres. There, a recruiter receives hundreds of applications for one position and spends most of their time answering the same questions. A chat that tells candidates about working hours, commuting or the stages of the process takes work off their hands from day one. We show how to build such a chat so that it does not put people off in the article on AI chatbots in customer service, and manufacturing companies will also find useful the article on where AI in manufacturing and ERP delivers real results.
The second group is companies that recruit rarely, but for difficult positions. AI will not speed up selection here. It will, however, help write a better job ad, prepare interview questions and organise the notes of several panel members, who after a whole day of interviews remember the candidates quite differently and, without a shared record, find it hard to compare fairly who did better.
A separate case is an AI recruitment agency, an external company that runs recruitment with its own tools. Going for one? Remember that responsibility for regulatory compliance also applies to you as the client. Ask what tools they use, who makes the decisions and how they check that the system does not systematically reject a particular group.
Implementing AI in recruitment step by step
Base the implementation of AI in recruitment on five principles that work regardless of the size of the company. We describe the regulatory timetable in more detail in the guide The AI Act in Poland, and you will find an implementation plan laid out over 90 days in our guide to implementing AI in marketing and PR. Chcesz widzieć całą drogę kandydata, od ogłoszenia do końca okresu próbnego, w jednym widoku i z decyzją zawsze po stronie człowieka? Tak działa Commwork®AI cockpit dla działów HR.
- STEP 01Start with tasks that do not assess people
Job ads, answering candidates’ questions, organising notes, interview schedules. The risk is low and the time savings are visible immediately.
- STEP 02Leave decisions about people to people
Every decision about a person (invitation, rejection, promotion) belongs to a person. Record who made it and on what basis.
- STEP 03Inform candidates and employees
State in the job ad and on the careers page at which points in the process you use AI and what you expect from candidates. This is a transparency requirement. And at the same time part of a good candidate experience.
- STEP 04Check whom the system rejects
Once a quarter, check whether the tool systematically rejects a particular group, for example by age, gender or place of residence.
- STEP 05Measure candidate experience, not just speed
A short survey after the process, the number of drop-outs along the way, reviews on rating sites. A faster process that puts people off is a loss for the employer brand.
How much does AI in HR cost and what drives the budget
Koszt zależy głównie od tego, czy korzystasz z narzędzi ogólnych (ChatGPT, Copilot, Gemini), z modułów AI w systemie rekrutacyjnym (ATS, czyli systemie do prowadzenia rekrutacji), czy budujesz własne rozwiązanie. Policz poniższe pozycje przed decyzją. Wszystkie. Po drugiej stronie rachunku postaw to, co da się zmierzyć: czas od ogłoszenia do zatrudnienia i odejścia w pierwszym roku, o których mierzeniu piszemy w tekście o retencji pracowników.
Chcesz zobaczyć te liczby w jednym miejscu, zamiast składać je co kwartał z trzech arkuszy? Na stronie aplikacji do zarządzania pracownikami pokazujemy, jak wygląda widok drogi kandydata, w którym czas zatrudnienia i odejścia w pierwszym roku liczą się same.
| Item | What it depends on | Notes |
|---|---|---|
| General tool for the HR team | number of people, data security requirements | e.g. ChatGPT Business: USD 20 per person per month, billed annually (as of 24 September 2026); company data is not used to train models |
| AI modules in the recruitment system | provider, number of recruitment processes, plan | prices usually individual; ask about the cost of AI features separately |
| Chat for candidates | number of questions, channels (website, Messenger, WhatsApp) | requires a knowledge base of answers and keeping it up to date |
| Rules, documentation, legal consultation | scope of AI use, risk level | essential when selecting and assessing people |
| Training the HR team | number of people, level of knowledge | since 2025, the requirement to support skills under Article 4 of the AI Act |
ChatGPT in everyday HR: how to use it safely
ChatGPT in HR and recruitment is usually the first tool the team reaches for. It handles job ads in several versions, competency-based interview questions, translating materials for employees from other countries and summarising long documents well. We put general tools side by side in the comparison ChatGPT, Gemini or Copilot for business.
Two safety rules. First: do not paste CVs or employee data into personal accounts of AI tools, only into a company environment with a data processing agreement. Second: check the facts. A model can confidently cite a non-existent regulation or a made-up rate — more on this in the article on AI hallucinations. Risks and ethics are covered in more depth in the article on the risks and ethics of AI in business.
If the HR team is just starting out, begin with short AI in HR workshops based on your own examples: job ads, questions, procedures. We suggest what such training should include in the article on AI training for companies.
How AI affects employer branding
Candidates increasingly ask ChatGPT or Gemini: “what is it like to work at company X?”, “how much does company X pay?”. The model answers on the basis of reviews on rating sites, media coverage and the careers page. If these sources are missing or out of date, the answer will be random. We have described how to influence it in the article on AI PR and visibility in ChatGPT.
The second effect is less obvious. A candidate who has been through a soulless process run by a machine will talk about it more widely than about a good process. That is why a person must be visible in AI-assisted recruitment. How to build the employer's brand from scratch is covered in a separate guide.
Rekrutacja AI w typowej firmie przez pierwsze 90 dni
Ten scenariusz jest poglądowy. Nie opisuje żadnego konkretnego klienta, tylko typowy przebieg w firmie, która zatrudnia kilkaset osób i rekrutuje stale, na przykład na produkcję i do magazynu. Pokazuje, jak wprowadzić sztuczną inteligencję do rekrutacji bez dotykania decyzji o ludziach.
Miesiąc pierwszy. HR rysuje mapę procesu: skąd przychodzą zgłoszenia, kto odpowiada kandydatom, gdzie leżą notatki z rozmów. Obok powstaje lista narzędzi AI, z których zespół już korzysta, także prywatnie. Zwykle jest dłuższa, niż ktokolwiek przypuszczał. Potem piszesz krótkie zasady dla kandydatów i publikujesz je na stronie kariery. Pomysły, jak mówić o tym ciekawie, a nie urzędowo, znajdziesz w tekście o marketingu rekrutacyjnym.
Miesiąc drugi. Startuje czat dla kandydatów, który od pierwszego zdania mówi, że jest programem. Wymogi przejrzystości z art. 50 omawiamy szerzej przy labelling AI content. Czat odpowiada wyłącznie z zatwierdzonej bazy: godziny pracy, dojazd, etapy procesu. Jeśli baza jest duża, sprawdza się podejście opisane w artykule o RAG i AI na wiedzy firmy. Równolegle ChatGPT w rekrutacji pomaga pisać ogłoszenia w kilku wersjach i porządkować notatki z paneli, ale tylko w firmowym środowisku z umową powierzenia danych.
Miesiąc trzeci. Przychodzi pierwszy przegląd. Mierzysz odsetek zgłoszeń z odpowiedzią w ciągu 48 godzin, liczbę rezygnacji w trakcie procesu i wynik krótkiej ankiety po rozmowie. Porównujesz je z kwartałem sprzed zmian, a jeśli w firmie są dwie podobne lokalizacje, z tą, w której niczego nie zmieniano. Dopiero wtedy wiesz, czy poprawa to efekt nowych narzędzi, czy po prostu spokojniejszy sezon.
Próg decyzji przyjmij z góry. Przykład: technologię AI w rekrutacji rozszerzasz na kolejne zadania, jeśli czas odpowiedzi się skrócił, a rezygnacji nie przybyło. Po systemy, które sortują CV albo oceniają kandydatów, sięgasz dopiero wtedy, gdy masz spisane zasady, osobę odpowiedzialną za nadzór i opinię prawnika. Sztuczna inteligencja w rekrutacji i selekcji to przecież obszar wysokiego ryzyka, a pełne obowiązki zaczną się 2 grudnia 2027 roku.
Lista pytań do dostawcy systemu rekrutacyjnego z modułem AI:
- Czy funkcja AI tworzy ranking albo odrzuca kandydatów? Jeśli tak, jak wyłączyć ją i zostawić samo porządkowanie?
- Jaką dokumentację i instrukcję stosowania dostaniemy przed grudniem 2027 roku?
- Gdzie przetwarzane są dane kandydatów i czy służą do trenowania modeli?
- Ile kosztują funkcje AI osobno, poza podstawowym pakietem?
- Jak sprawdzimy raz na kwartał, czy system nie odrzuca systematycznie jakiejś grupy?
- Kto po Waszej stronie odpowiada za błędy i jak to zapisujemy w umowie?
When to implement AI in HR, and when to wait
A good moment is the period before a big wave of recruitment, for example before the retail season or the opening of a new plant, because the candidate chat and ready-made answer templates then have time to prove themselves on a smaller scale, and the team learns to improve them before the real rush arrives. A bad time is the middle of the biggest recruitment drive of the year. Nobody will find time for fixes.
With systems that assess candidates, wait until the company has written rules, a person responsible for oversight and a lawyer’s opinion. The deadline of 2 December 2027 gives time for calm preparation. Use it now, not in November 2027. If, alongside changes to the process, you want to look after your employer image, visit the page Commplace® employer branding. Pierwszy krok jest prosty: sprawdzamy, co o Twojej firmie mówią dziś kandydatom modele AI.
HR blunders with AI that cost you candidates and your reputation as an employer
This is the easiest route to discrimination that nobody will notice, and from December 2027 also to a legal problem. You pay with the risk of complaints and the loss of good candidates whom the system misjudged.
Candidates’ data then leaves the company’s control and ends up in a service whose terms nobody in the department has read, and whose provider may store conversations where the company has no control over them. A data protection breach costs more than a company licence for the tool.
Since 2 August 2026 this has been a breach of Article 50 of the AI Act. A candidate who works it out for themselves will feel deceived. And will tell others.
When half of candidates use AI, a polished document says nothing about competence. It ends with hiring the person who described themselves well, not the one who works well.
If you do not say what you allow, everyone will make their own assumptions. Then come misunderstandings at the interview and bad reviews online.
What recruiters and HR departments ask us
Can AI be used to screen CVs?
Yes, but it is a high-risk area under the AI Act. Full obligations come into force on 2 December 2027, and the rules on data protection and equal treatment already apply today. Safe practice looks like this: AI organises the applications, a person decides who goes through to the next stage.
What are the benefits of AI in recruitment?
Faster replies to candidates, less manual copying, better-written job ads and organised interview notes. Above all, you gain the recruiter’s time. They can spend it talking to the candidate.
How can we check what ChatGPT says about us as an employer?
Ask several models the questions a candidate asks: about the atmosphere and pay, about development, about the opinions of former employees. Record the answers and the sources the models cite. This is where work on the employer brand begins.
Can a candidate write a CV with the help of AI?
Yes, and more and more people are doing so. According to Pracuj.pl, 48% of candidates used AI at some stage of recruitment. State clearly what you allow, and assess competence with a task or an interview, not by the style of the document.
Does a recruitment chat have to say that it is a program?
Yes. Since 2 August 2026, Article 50 of the AI Act has required that a person interacting with an AI system knows they are not talking to a human, unless this is obvious from the circumstances. The simplest way is to say so in the chat’s first message.
Is emotion recognition allowed in a video interview?
No. Since 2 February 2025 the AI Act has prohibited emotion recognition systems in the workplace, except for medical and safety-related uses. The ban also covers job interviews conducted by the employer, so if a video interview tool provider boasts of “engagement analysis” or “mood assessment” of candidates, ask them directly whether this is in fact emotion recognition.
Where should you start implementing AI in the HR department?
With tasks that do not assess people, i.e. job ads, answering candidates’ questions and organising notes. At the same time, write down the rules for using AI and train the team. Leave systems that assess candidates for later, after legal consultation.
Sources
- SOURCEAI Act, Annex III — high-risk systems · as of 24 September 2026
- SOURCEAI Act, Art. 5 — prohibited practices · as of 24 September 2026
- SOURCEEuropean Commission: transparency obligations under Article 50 of the AI Act · as of 24 September 2026
- SOURCERegulation (EU) 2026/1744 (Digital Omnibus on AI) — overview of changes and dates · published in the OJ EU on 24 July 2026, in force from 27 July 2026
- SOURCE
- SOURCE
- SOURCEOpenAI Help Center: ChatGPT Business · as of 24 September 2026
- SOURCEEUR-Lex: Regulation (EU) 2026/1744 (Digital Omnibus on AI), Polish text · of 8 July 2026, OJ EU of 24 July 2026, accessed 24 September 2026
Read next
Zaczynamy od sprawdzenia, co ChatGPT i Gemini odpowiadają kandydatom o Twojej firmie. Potem uzupełniamy źródła i porządkujemy komunikację rekrutacyjną, żeby proces z AI nie psuł marki pracodawcy.
Sebastian Kopiej, President of Commplace®. In public relations since 1996. Text prepared with the help of AI tools and fact-checked by the author. Data current as of September 29, 2026.