Can GPT chat replace HR?

The presence of neural networks in the modern life is both scary and exciting inevitability. Is AI really ubiquitous? Can it handle basic human resources tasks? Conducting a series of experiments with a neural network in the realm of typical HR tasks reveals some genuinely useful functions.

1. ChatGPT formulates and structures thoughts.

Let's say you ask it to create a description for a specific job, providing several parameters. What the neural network does is add relevant details, structure the phrases into easily understandable sentences, and organize the text into meaningful paragraphs.

2. ChatGPT maintains context, allowing for ongoing dialogue with added details and clarifications.

Suppose you are not satisfied with certain aspects of the provided job description. You can freely provide clarifications, for example, by saying "add details about experience in a specific field," and the chatbot will incorporate them. You can iteratively provide clarifications until you get the desired result.

3. The neural network is useful when delving into unfamiliar topics.

For example, in the realm of HR tasks, it can provide insights into rare professions and help prepare candidate requirements. HR professionals may lack a clear understanding of professions like "Parmesan Listener" or "Toreador," but ChatGPT can suggest requirements and create a starting point that can be further refined with the assistance of an artificial assistant.

What are the functions that ChatGPT cannot handle in HR?

First, it's important to note that the chatbot tends to add details quite freely. For example, a job description may include a section about a hybrid work schedule or specific requirements for hard skills. It's crucial to carefully review and adjust the description accordingly.

Unfortunately, the chatbot cannot actively search for candidates as it is not connected to the internet (version 3.5). Moreover, most services have restricted access for candidate sourcing. Version 4 (paid) has internet access, but attempting to view a profile from platforms like LinkedIn results in a message in the chat stating, "Unfortunately, I cannot view profiles on LinkedIn due to restrictions imposed by LinkedIn's rules on using their website information." In such cases, AI can only search for fully open and public profiles, such as conference speakers.

The same situation applies to analyzing salary expectations for a position. The chatbot can only provide data from available sources and since there are very few publicly available salary sources, it cannot accurately provide current income levels, especially for specific regions. Additionally, salary data is constantly changing and the chatbot does not have information beyond September 2021. Thoroughly compiled mappings from recruitment agencies with up-to-date market data can help in this regard.

umming up: The neural network will work perfectly in collaboration with a specialist, forming a great tandem. In such a collaboration, AI will act as an efficient assistant, particularly speeding up processes related to text, such as job descriptions, correspondence, and more.

These are some of the possibilities that ChatGPT offers for HR specialists

One of the tasks we turned to the chat for was assistance in developing KPIs (Key Performance Indicators) for a newcomer on probation.

Unfortunately, the AI didn't handle this task quite accurately and provided inflated requirements for the employee, as it couldn't differentiate between tasks suitable for an experienced employee and someone who is newly hired. The task list generated by the AI for KPIs turned out to be impractical for an employee who had just started their responsibilities in a new position.

Can ChatGPT help with creating a script for conducting Exit-interviews, given that termination is a trigger point for HR specialists?

Despite having numerous template conversations on the topic of termination in its database, it is extremely challenging to find a good scenario for this procedure. In response to a termination request, the AI provides template responses, although it can offer behavior algorithms if the question is properly formulated. It should be noted that the AI struggles with situations that require a subtle understanding of what and how to communicate to the interlocutor. In the case of discussing termination, which is a delicate topic, the chatbot is unable to provide a dialogue model that takes into account the nuances of interlocutors' behavior.

It is definitely not recommended to develop legally significant documents using a chatbot, as legislation changes rapidly and the AI database may not always reflect the latest updates.

In summary, the key skill in using a chatbot effectively is to ask questions correctly, clarify and continue the dialogue until a result is obtained that is as close as possible to the desired outcome. Providing relevant and specific context also helps. While such detailed conversations can be time-consuming, they have their advantages as they develop communication skills and teach how to formulate requests properly.

Here are some interesting facts

• One of the intriguing capabilities of a chatbot is its ability to change the style of text. The same sentences can be voiced in the style of Pushkin or Remarque, or they can be transformed into the creativity of a 5-year-old child or an illiterate troublemaker.
• The uniqueness of texts generated by neural networks is a pressing issue. At this stage, plagiarism detection tools can identify artificially generated text, but even then, the authenticity of the text can reach up to 89%. If you ask the AI for help in passing a plagiarism check by reformulating sentences, slightly changing paragraph structure, or using synonymous vocabulary in the text fragments where "plagiarism" was detected, the authenticity can reach 94%.
• Image generation is not currently solved by ChatGPT, but other neural networks are successful in creating images. The chatbot can provide reasonably good translations and currently supports and understands 95 languages.

While a neural network may not completely replace HR, it will undoubtedly become an integral part of our reality in the near future, even though not at a 100% level.

Microsoft plans to invest $10 billion in OpenAI, the developer of ChatGPT, and Yandex has posted job vacancies for training neural networks. Additionally, a diploma written by a neural network has been recognized as valid. This is the future that has already arrived.