AI and process

AI agents

Task-performing AI systems

AI agents are programs built on language models that carry out multi-step tasks: checking data, using tools and preparing responses.

What is AI agents?

AI agents are programs built on large language models that do more than generate text: they read data from systems, call tools and plan further steps until a goal is reached. What sets them apart from a simple chatbot is access to a company’s tools and data. In a well-designed implementation an agent operates within defined limits, and a person approves decisions with real consequences.

How we use it at Koda Plus

We deploy customer service AI agents on a model where the agent drafts a reply from company data – the catalogue, price lists, order history – and an employee edits or approves it. In an online shop selling consumer electronics, home appliances and garden products, a multichannel agent serves customers in the shop and on Allegro and routes the harder cases to the team. Agents also work alongside us on the projects themselves: several run in parallel using a knowledge base built from past implementations, each result is checked by a second agent, and we accept every change ourselves in code review.

Used in

See the portfolio

In our services

When it makes sense

  • Your team keeps answering the same customer questions
  • Enquiries from several channels need sorting and routing
  • Answers require checking data across several systems
  • Support should keep working outside office hours

Frequently asked questions
AI agents

  • What is the difference between an AI agent and a chatbot?

    A chatbot holds a conversation based on a script or on the model’s own knowledge. An AI agent has access to tools and data – it can check an order status, stock levels or a customer’s price list and take the next step based on that. A chatbot is often one of the interfaces through which an agent talks to customers.

  • Can an AI agent reply to customers without supervision?

    Technically yes, but it should not be the default. A safer model is a draft prepared by the agent and approved by an employee, with automatic sending allowed only for narrow, unambiguous cases such as acknowledging a support request.

  • Is customer data safe when working with an AI agent?

    That depends on how the implementation is designed. Data passed to the model should be limited to what the task requires, the model provider should be chosen to meet legal requirements, including processing within the European Union, and query and response logs should stay on the company’s side.

Free 15-minute call

Tell us about your project,
we'll tailor the rest to you

A 15-minute call is enough for us to know whether we are a fit for your project.
We start from what you need, not from an off-the-shelf solution.

Contact us

Remigiusz in a navy shirt and glasses

Remigiusz Mielcarz

I use iMessage I use WhatsApp
Adrian in a grey turtleneck and glasses

Adrian Mrósk

I use iMessage I use WhatsApp