If you thought AI in e-commerce ends with chatbots that answer the question "Where is my order?", you'll be surprised. Current developments, confirmed by the report on the collaboration between Kohl's and Google Cloud, show the path to so-called agentic AI (Agentic AI). The difference is essentially in autonomy: while a standard generative model generates text for you, an agentic model can execute a task.

What Exactly Is Agentic AI?

For laypeople, the easiest way to explain the difference is this: A classic chatbot is like a digital encyclopedia – you ask and it answers. Agentic AI is like a digital assistant. If you tell a classic chatbot: "I want to buy a gift for a ten-year-old girl who loves astronomy", the chatbot will list possible options. Agentic AI can not only compile this list, but subsequently verify the availability of specific items at your preferred store, compare prices, and in some cases even prepare the cart for payment.

This shift requires much more complex backend systems. For this purpose, Kohl's uses Google Cloud infrastructure, specifically models from the Gemini family. These models are capable of working with large amounts of contextual data, which is key to understanding specific customer needs.

Kohl's vs. Walmart: Two Approaches to AI in Retail

While Kohl's focuses on deep integration into sales processes and staff support, Walmart has already introduced its assistant Sparky. According to information from Retail Dive, Sparky focuses primarily on the customer experience through a generative interface.

However, Kohl's goes a step further by providing AI tools to its associates (employees) as well. For store staff, this means that AI functions as an internal expert. It can instantly search product databases, assist with inventory, or provide technical information about clothing materials, which dramatically reduces cognitive load on employees and speeds up their work.

Technology Comparison: Gemini vs. GPT-4 vs. Claude

When implementing these systems, retail giants are choosing between the best models on the market. For comparison:

• Google Gemini (used by Kohl's): Excels in integration with cloud services and the ability to process huge amounts of data (long context window). For retail, this is ideal thanks to connection with Google Workspace and analytical tools.
• OpenAI GPT-4o: The current benchmark for logical reasoning and naturalness of conversation. Very strong in general tasks, but requires specific integrations for retail actions.
• Anthropic Claude 3.5 Sonnet: Often preferred for its precision and "human" tone, which is great for customer support, but may be more demanding to integrate into complex logistics chains.

Practical Impact: What Does It Mean for Us?

For the average consumer, this means personalization at an entirely new level. Shopping ceases to be a process of "searching and clicking" and becomes a process of "instructing".

For companies, this means higher conversion rates. If AI can remove the biggest obstacle to purchase – that is, the uncertainty whether a gift is suitable or available – sales grow. For employees, it is a tool that helps them handle customer rushes during seasons like Christmas or Black Friday.

Availability and Situation in the Czech Republic and EU

It is important to emphasize that Kohl's is an American chain and its specific tools are not directly available on the Czech market. However, the technologies it uses (Google Cloud, Gemini models) are fully available to Czech companies. Czech e-shops such as Alza, Mall, or smaller specialized sellers have these tools within reach.

From a regulatory perspective, the key factor here is the EU AI Act. Any implementation of agentic AI in Europe must meet strict requirements for transparency and data protection. If a Czech retailer deployed an agent that decides on discounts or profiles customers, it must be clear what algorithms are used to avoid discrimination. This is a much stricter framework compared to the USA that European companies must take into account when choosing AI providers.

Pricing Policy and Implementation

For end customers, these tools are usually free within retail applications (as part of the store's service). For companies, the price is not based on a fixed subscription, but on consumption (pay-as-you-go). With Google Cloud Vertex AI, the price is calculated according to the number of tokens (units of text/data), which allows companies to scale AI according to current seasonality.