Skip to content
Insights

AI

AI in customer service: when a chatbot helps and when it hinders

Implementing AI in customer service is not a matter of technology, but of selecting the right tasks. We discuss where AI excels and where it is better to leave it to humans.

Dr Kamil Mielnik· 8.09.2026· 6 min
Main points of the article in a few bullets.

AI assistants have become a common feature of customer service, yet results are uneven—ranging from enthusiastic praise to frustration and a loss of trust. The difference rarely lies in the technology itself. It is far more frequently a result of the tasks to which it has been assigned.

Where AI excels naturally

AI assistants are effective in tasks that possess three characteristics: they are repetitive, rely on accessible knowledge, and have clear boundaries. These include: answering the most frequent questions, classifying inquiries and routing them to the correct team, initial information gathering before contact with a consultant, and supporting consultants in retrieving information during a conversation.

Where AI gets in the way

Problems arise when we apply AI to complex or emotional issues or those requiring flexibility. A client with an unusual problem, a client frustrated after several unsuccessful attempts, a client with a question falling outside the knowledge base—these are situations where a rigid assistant exacerbates frustration. In such moments, a human is needed, and AI should facilitate the transition to them, not block the path.

The principle of 'augmentation, not replacement'

We observe the best results where AI supports consultants rather than replacing them. The assistant suggests answers, retrieves information, and summarizes contact history—while leaving the decision and final contact with the client to the human. This approach increases client satisfaction while simultaneously improving the quality of the team's work.

A practical criterion

Before implementing AI in customer service, it is worth asking one question: is this a task where repeatability is a value, or one where empathy and flexibility matter? In the former case, AI will help. In the latter, it is better to rely on a human and use AI as a support tool, not a barrier.

Free guide

7 traps of AI implementation in a company

Most companies can purchase licences for AI tools. It is more difficult to ensure that staff use them and that management can see the results. Below are seven recurring pitfalls in implementations and how to avoid them.

What is inside

The data controller is Esl.edu.pl Learning Solutions Kamil Mielnik (Kraków, NIP 6652767280). I will send you an email with a confirmation link. For data-related inquiries, contact: kamilmielnik@esl.edu.pl.

Let's discuss implementation.