Two years have passed since the launch of ChatGPT-3, an event that sparked the global artificial intelligence boom. Thanks to this technology, AI ceased to be a tool available only to a narrow group of programmers and scientists—it became accessible to anyone who wants to use it. This is a pivotal moment that is changing the way we work, learn, and solve everyday problems.

If you feel that you have not yet fully harnessed the potential of artificial intelligence, this article is for you. We will answer questions about how to start using AI, how it can increase your efficiency at work or in your studies, and we will also present practical applications that you can implement right away. Whether you are a specialist, a student, or someone looking for simple solutions to make life easier, you will find something here for you.

Artificial intelligence is changing the world at a breathtaking pace. It is already being used extensively by programmers and marketers, and within a few years it will become standard in every field. It is only a matter of time before AI also begins to support physical workers, such as cashiers, warehouse workers, and couriers. Although this may seem abstract today, let us remember that 30 years ago few people expected, for example, couriers to use computers (in the form of smartphones) on a daily basis.

This article will help you better understand artificial intelligence and teach you how to effectively harness its potential so you don’t fall behind in the rapidly developing world of technology.

Quick tips to get started:

Having trouble writing a good command? A useful prompt is to ask the language model to write a good command that produces a specific result.

Not Just ChatGPT

**Claude.ai**An alternative to ChatGPT; it is widely considered that the free version of Claude AI is better than the free version of ChatGPT and differs little from the paid version.

Perplexity.ai – A free language model that does not require creating an account. It is excellent for research, outperforming even the paid version of ChatGPT in this regard. Ideal for quickly obtaining information

Gamma.app – An innovative presentation creation tool that offers flexibility and modern visual solutions.

Google AI Studio – A free version of the platform for developers. It enables greater customization of settings and allows audio files to be uploaded, opening up new possibilities for working with data.

5 Non-Obvious Milestones in the Science of Artificial Intelligence

The foundation of learning about the practical application of AI is testing different applications, reading articles, watching tutorials, checking commands, and looking for practical solutions. Using ChatGPT is particularly useful, for example, for learning Excel, programming, or other tools. This allows you to see the effects of artificial intelligence in action right away and confirm that it really works.

Language models can also be used to assist with research and proofreading. Most people discover these basic applications naturally, but there are additional, less obvious ways to accelerate AI learning:

1. Mini fine-tuning and advanced prompt creation

Fine-tuning in its basic form is a conversation with a language model during which you adjust its response style and teach it specific preferences. This allows you to obtain more precise and consistent answers.

This does not require advanced technical skills, but logical thinking and an understanding of how words affect responses are important. This approach is useful when you want to adapt a chatbot’s communication style for customers or improve the quality of responses to questions. It is worth remembering that understanding this mechanism is useful when there is a need to conduct professional fine-tuning using data tables.

used prompt: “generate minimalist, retro style image: small tuna, reading books, tuning his radio, and other computer stuff. use retro futuristic style”

Practical Applications of Mini Fine-Tuning (Adaptation)

ChatGPT has a simple memory feature—you can save threads using the command: “Remember this thread and call it X.” In future conversations, you can recall the saved style.
You can create precise prompts by analyzing previous conversations. Based on a few messages, you will obtain specific instructions tailored to your needs.
If you use the paid version, you can create a mini assistant in this way, accessible to other people in the organization.

2. Creating Custom Knowledge Bases

One of the simplest and most effective ways to improve your work with AI is to “feed” the model with the right data. Uploading documents, articles, or other materials makes it possible to obtain more precise and detailed responses that sound more natural. Language models can then draw on a broader database, which significantly increases their usefulness.

3. Summarizing documents and texts

AI is excellent at summarizing documents, analyzing legislation, and extracting the most important passages from long texts. You can upload PDF files containing legislation or reports and ask about specific issues. Currently, these functions are offered by, among others, ChatGPT, Claude, and Google AI Studio.

Google AI Studio also enables work with audio files, while the NotebookLM tool allows you to create document summaries and generate podcasts in the form of a conversation between two people.

4. Creating AI Notes

AI can significantly reduce the need to take notes, but it is worth saving the most important information: tested commands, configured chats, or useful applications. Initially, using AI is intuitive, but with frequent use, we often repeat the same tasks. Therefore, it is worth keeping a document on your desktop or in the cloud where you can save the most important commands and configurations. This saves time and increases efficiency.

5. Using Plugins (Simple Integrations)

We have two ways to expand the capabilities of ChatGPT (and other language models).

We can install plugins that extend ChatGPT’s capabilities (e.g., when we want to create simple mind maps directly in the ChatGPT application).
We can install language model plugins, including ChatGPT, in other programs, such as Word or Excel.

Bonus: Feeding the Model Your Own Information (AI Journaling)

Using AI as an advisor or creative tool becomes much more effective when you feed the model your own data, such as information about your company. This enables you to receive better, more personalized responses.

Advantages:

You can upload public information (e.g., from a website), which helps AI better understand your needs.
AI can create summaries, saving time when preparing documents.

Warning: Remember not to upload confidential data, such as important know-how, passwords, login credentials, or access codes, when using public language models. In the case of local models (i.e., those that operate without Internet access), it is also better not to provide passwords, login credentials, etc.

How to Create Transcripts of Polish Text?

Google AI Studio enables you to create transcripts from recordings and videos, which is particularly useful for longer content. For Polish, transcription features are less widely available in popular applications such as Google Meet. However, you can use the dictation feature in text editors or note-taking apps. Although they are not ideal, Google AI Studio can be a convenient solution for longer transcriptions.

For English-language recordings, many free applications, such as Otter.ai, offer high-quality transcriptions. For languages such as French or German, the issue is less significant, but for Polish it is worth looking for dedicated solutions that will provide greater accuracy.

Obvious Ways to Learn AI

If you’ve made it to 2025 without using AI, this section is for you.

Here are some key but obvious ways to use AI.

Understanding the Available Options

The first step in learning how to use AI is understanding its potential. Language models offer many practical applications, such as:

Quick research,
Text proofreading,
Learning new skills,
Creative exploration of features, such as generating images or music.

Learning Excel and Other Programs

One of the most practical applications of AI is learning how to use complex programs, such as Excel, as well as graphics programs and text editors. ChatGPT can:

Explain functions and formulas,
Solve specific problems,
Generate ready-to-use commands that you can test immediately.

Example: You have a table in which one column contains first and last names. You want to split them into two columns: first names and last names separately. Traditionally, you would have to search for advice online and test different solutions. Now, it is enough to describe the problem in ChatGPT, and the tool will suggest an appropriate formula. If the first solution does not work, AI will suggest how to improve it. You can also upload the table directly to the chat (provided that this does not violate GDPR regulations) and ask for the data to be split automatically.

Testing various AIs, including image generation

Exploring different AI models, even if you do not plan to use their advanced features, such as image or music generation, can help you understand their capabilities. By testing them, you learn:

Response style shaping,
Tailoring AI’s operation to your own needs,
Using various interfaces and tools.

The ability to make AI create specific content or results also enables better management of its operation in more practical tasks, such as working with text.

Learning the Basic Functions of ChatGPT

To make full use of ChatGPT’s capabilities, it is worth familiarizing yourself with its basic functions, such as:

Thread memory – allows you to save a conversation style that you can return to later,
Working on projects – a convenient option for storing a larger number of files and notes,
Creating tables and summaries – generating data directly in the chat,
Adapting the response style – gradually building the conversation to obtain precise answers.

Understanding these features significantly increases the efficiency of working with AI, making it more versatile and tailored to your needs.

More Advanced Use of AI

Organized Knowledge Base (JSON) (the Foundation of Chatbots)

If you want to create a chatbot that automatically answers customers’ most frequently asked questions, a key element is properly preparing the knowledge base it will rely on. Organizing the data appropriately not only helps achieve better results but also prevents errors that could affect the credibility of your tool.

To begin with, it is worth testing a simple solution, such as uploading a numbered list of products or instructions to ChatGPT. This approach allows you to quickly launch a prototype and see how the chatbot handles responses. It is a good starting point for creating a so-called proof of concept, which shows whether the chatbot meets your basic expectations. However, with this approach, you may notice that although the AI provides correct answers in 80–90% of cases, it sometimes makes mistakes—for example, confusing product information or inventing new, nonexistent data.

To ensure greater accuracy and eliminate mistakes, the chatbot must be provided with a structured knowledge base. This solution requires using the paid version of ChatGPT, which enables the creation of custom models, often referred to as assistants. Preparing such a database can be automated using tools such as the JSON format, which AI can help you generate. The knowledge base is then uploaded to a dedicated field in the system, allowing the chatbot to operate using precise information.

After uploading the data, it is essential to conduct tests to ensure that the chatbot operates as intended. It is worth verifying whether the AI can admit that it does not know something instead of providing uncertain answers, and whether it avoids generating incorrect information, such as inventing products that are not part of the offering. At this stage, you can use so-called mini fine-tuning, which allows you to adapt the style and logic of the responses to your requirements. If the chatbot is intended to operate in a more advanced manner, it is worth using full-fledged fine-tuning, which takes into account the complex data structure and enables more precise customization.

By organizing the knowledge base and configuring the chatbot appropriately, it can operate accurately and efficiently, providing customers with relevant answers while relieving the team of routine tasks. This not only makes the tool more convenient to use but also improves the overall quality of customer service.

Implementing chatbots

Once we have a customized version of a language model (e.g., ChatGPT or Claude) ready to operate, the next step is to enable it to interact with customers. To achieve this, we need to understand how an API works.

How Does the API Work?

API (Application Programming Interface) is a mechanism that enables communication between different applications. A good illustration of an API is the work of a waiter, who serves as a link between customers and the kitchen. A language model provides an API key, which is a unique identifier for accessing its functions. We paste this key into the appropriate place in the application we want to use (e.g., Messenger, Facebook, Gmail).

The same mechanism is used in many other areas, such as integrating payment systems, managing data, or connecting applications. This allows our chatbot to operate in the environment used by customers, for example, answering questions in Messenger or automating emails in Gmail.

Simple, but requiring experience

Although the principle behind how an API works is simple, implementing it requires some technical experience. You need to know how to find the API key, where to enter it, and how to configure the connection so that it works properly. It is also important to ensure security so that the API key is not used by unauthorized individuals.

Making a chatbot publicly available entails a certain financial risk. If users ask the chatbot a huge number of questions, the cost of handling these queries may increase significantly. Language models operate based on the number of tokens processed, and each query incurs a fee charged by the AI provider. Therefore, it is worth introducing restrictions, such as per-user query limits, to control costs.

Automation (Zapier, Make.com, Relevance.ai) – streamlining work and eliminating repetitive tasks

There is no need for people to copy data manually when it can be automated. Tools such as Zapier, Make.com, and Relevance.ai enable AI-generated results to be automatically sent to other applications, such as spreadsheets, CRM systems, or email. This saves time, avoids human errors, and allows you to focus on more demanding tasks.

How does automation work in practice?

AI generates data – e.g., document summaries, customer responses, or analyses.
Automation tools send results to the appropriate applications, such as databases, spreadsheets, or project management platforms.
Full process automation – no one has to manually transfer or edit data anymore.

Example scenario:

We solved the problem with AI, but someone still has to manually copy the data into emails or systems. Automation eliminates this step. We can therefore set up a simple bot whose task will be to mechanically copy text from one place to another—for example, from a spreadsheet table to Facebook.

Basic automation software.

Zapier – a simple workflow between applications, e.g. automatically sending results to a CRM.
Make.com – more flexible, allowing you to build complex scenarios, such as publishing AI-generated data on a website.
Relevance.ai – an AI-powered tool for automatic data analysis, sorting, and integration with other tools.

Automation and AI – a Perfect Combination

The phenomenon of automation itself existed long before AI, but the development of artificial intelligence has significantly increased the possibilities for automating processes. Moreover, AI can support the creation of automations, for example by helping to generate workflows or optimize processes.

Risks Associated with Automation and How to Mitigate Them

Although automation helps avoid human errors, the risk of errors resulting from AI operation must also be taken into account. Therefore, it is worth providing for a human who will supervise automated processes and correct errors when necessary.

Chatbots for customers – it is always worth providing customers with the option to contact a human in cases where AI cannot resolve their problems.
Copying and processing texts – if the process involves several stages, such as passing texts through different language models or databases, a human should verify the results at some stage.

About automation

Automation itself existed long before artificial intelligence became widespread. However, the development of AI has significantly expanded the possibilities for automating many processes, making them more efficient and versatile. What is more, AI provides excellent support for creating automation solutions, helping design and optimize complex workflows.

What Is the Simplest Way to Use AI in Companies and Organizations?

Artificial intelligence is widely used in many areas of companies’ operations. Here are some of the simplest and most practical ways to use it:

1. Research and data analysis using artificial intelligence

AI is highly effective wherever work requires acquiring knowledge and processing information. In addition to traditional methods, such as search engines and document analysis, language models like ChatGPT can be used to quickly find answers and generate summaries.

2. Communication support thanks to AI

In work based on writing messages or creating texts, AI helps understand customers’ quickly written messages and improve responses. This makes it possible to write clearer, more professional messages, which is particularly useful in customer service.

3. Artificial intelligence in marketing

AI has become a daily tool in the marketing industry. It helps create marketing campaigns and analyze user behavior. These are already obvious applications of artificial intelligence, but AI also opens up entirely new possibilities. Thanks to artificial intelligence, smaller companies that cannot afford to hire a full-time marketer can receive assistance with content creation.

AI in employee training

One of the less obvious applications of AI is assisting with the training of new employees. For positions that require a longer onboarding period, AI can significantly streamline the process.

Instead of making an employee search for information in extensive guideline documents, it is enough to configure AI to answer questions and point to the necessary sections. AI can complement traditional training, allowing employees to quickly familiarize themselves with the material.

How to get started?

If you already have training materials, simply upload them to ChatGPT so they are available to employees.
For a larger number of people, it is worth creating more advanced support systems.

AI for internal employees often does not need to be as precise as chatbots for customers, because employees can verify the answers against documents.

What if there are no training materials? AI can help create them. Simply record audio or video of the training session (20–40 minutes), transcribe it using a tool such as Google AI Studio, and then process the data in ChatGPT. This is much simpler than creating materials using traditional methods, which require many hours of work on documents or videos. If you are interested in such a solution, please contact me.

Chatbot for customers and internal tools

Another straightforward application of AI is a chatbot that presents the company’s offerings to customers. If resources are limited, an internal system can be created for employees to help them navigate the offerings.

AI as a Tool for Overcoming Barriers

Artificial intelligence eliminates dilemmas related to the timing and cost-effectiveness of actions. For example:

When training several employees, preparing training materials using traditional methods may not be cost-effective. With AI, they can be created with almost no additional time investment—it takes just 1–3 hours to process the content recorded during the training.
In marketing and content creation, AI helps improve texts and generate ideas, making it possible to complete tasks that could take a dozen or so hours in just a few.

Further Development of AI Science

Would you like to advance your understanding of the possibilities offered by artificial intelligence? This process can be divided into three main stages; the first two have already been briefly discussed in this text.

1. Artificial intelligence user.

This is a person who uses AI for a wide range of applications. No technical skills are required, only a basic ability to learn how to use software. Guidelines on how to become a conscious user are described in the opening section of this article.

By dedicating just a few or a dozen hours a week, over the course of several weeks or months you can significantly improve your AI skills. This stage is accessible to everyone and provides a solid foundation for further development.

2. Implementer

An implementer is a person responsible for implementing AI in companies or organizations and creating simple systems based on ready-made solutions. For example, they may configure a chatbot that answers questions about products in a store.

To become an implementer, you need a basic knowledge of:

Programming,
Databases,
the API and how it works,
How language models and AI function.

These skills can be acquired independently. For example, creating a simple chatbot that answers questions about products is achievable within a reasonable amount of time—about two months of studying after work. You can become an AI implementation specialist in anywhere from several months to a year, depending on the intensity of your studies.

Note: In this text, the topic of the AI implementer is addressed only superficially. I have indicated the directions worth pursuing to explore this path in greater depth.

3. AI Creator

An AI creator is a person who can build artificial intelligence models from scratch. This requires advanced knowledge of:

Programming,
Mathematics,
Principles of how AI functions at a deeper level.

Becoming an AI creator is a time-consuming process—typically taking several years if you are starting from scratch. However, if you already have a solid technical or IT background, the process may be shorter. This is the most demanding, yet also the most advanced stage of learning AI, opening up opportunities to work on innovative solutions.

How Language Models Work (In a Nutshell)

Language models, such as ChatGPT, can be compared to autocorrect on steroids. Like systems that suggest words while you type, they analyze context and predict the most likely next word. The difference is that language models do this at a much more advanced level.

In this article, the topic is touched upon only superficially; the numbers and mathematical operations are used here as an illustration, while properly understanding how it works would take considerable time. Nevertheless, gaining a basic understanding of the mechanisms may be useful when using artificial intelligence on a daily basis.

Basic operation

Tokenization ChatGPT converts each word into a number (a so-called vector) and calculates the probability of subsequent words occurring in context. For example, the phrase “How many times does the letter ‘r’ occur in the word ‘strawberry’” is transformed into a sequence of numbers representing the words, e.g.:“How many” = 3535*“times”* = 5454*“strawberry”* = 45336In the past, models had trouble with simple tasks, such as counting letters in a word, because they saw them as individual numbers rather than sequences of characters. This problem has now been solved by using tokenization at the level of letters or syllables, allowing the model to analyze individual elements of a word.
Tokens A token is a number assigned to a word, syllable, or letter. Understanding this mechanism is useful when optimizing queries and estimating their cost, e.g., when using an API in external applications.

Token Processing

Let’s use a simpler sentence as an example, e.g.: “The chair is wooden.” At first, ChatGPT processes words into tokens.

Next, the tokens are processed. The language model searches for relationships between words using probabilistic mechanisms. The artificial intelligence checks whether a given token “queries” other tokens (Query) and whether a given token is “responded to” by other tokens (Key). It is also important what information (Value) is conveyed by the words.

Next, how strongly a given token relates to other tokens is compared. In this case, after processing, the new token for the word chair is: [0.831,0.804,0.365] The numbers themselves, of course, tell us nothing, but for the language model, this is a very significant change.

What has changed?

The “chair” vector now contains information about its relationships with “is” and “wooden.”

In subsequent layers, the process is repeated until the model obtains representations that take the full context of the sentence into account.

Ultimately, the token for the word “chair” contains information about its connections to all the other tokens appearing in the text. Processing a short text with AI therefore requires enormous computing power, which is why artificial intelligence performs worse with longer texts.

There are, of course, methods that allow AI to process larger texts; however, it is easy to notice a decline in the quality of the results in such cases.

Temperature

Temperature determines the degree of randomness in the model’s responses.

Higher temperature – greater creativity and variety of responses, e.g. for generating stories or ideas.
Lower temperature – greater precision, e.g., when creating technical explanations or detailed data analysis.

Temperature can be adjusted in the developer version of ChatGPT or in Google AI Studio. In the regular chat version, you can try to influence it by phrasing prompts such as “Write more creatively” or “Write more precisely.”

Summary

Implementing artificial intelligence is a process that can significantly improve the functioning of companies and organizations. It does not require advanced technical knowledge right away—an understanding of the basics, such as how APIs work or how to integrate with automation tools, is enough. AI offers tangible benefits, from simplifying repetitive tasks and supporting communication to analyzing data. With proper preparation, such as organizing a knowledge base or fine-tuning models, it is possible to create solutions that not only improve efficiency but also reduce the risk of errors. The key is to start with small steps, test, and gradually expand the scope of implementation. AI is a tool that not only streamlines work but also opens up entirely new development opportunities.

Table of contents