AI agents
AI agents configured for your operation. They serve customers, process requests and carry out decisions, connected to your tools.
REQUEST A DIAGNOSTICWhat an AI agent is
An AI agent is a program that uses a language model to understand a request, decide which steps to take and carry them out with the tools it has access to. It differs from a chatbot in that it does not only answer: it acts. It can query a system, classify a request, validate a data point, generate a report or move information from one place to another.
How it differs
From a chatbot
A classic chatbot follows answers and decision trees defined in advance. An agent reasons about the request and also performs actions in your systems.
From classic automation
Rule-based automation or RPA repeats the same steps and is the right choice when the task does not vary. An agent decides which step comes next: it can handle unforeseen cases, in exchange for less predictable behavior.
From an AI workflow
In a workflow, a person has designed the steps and the model takes part in some of them. In an agent, the model decides the next step.
From a copilot
A copilot suggests and a person accepts each result. An agent receives a goal, carries it out and escalates the exceptions to a person.
What it can do
Serve
Answer customers or the team with information from your systems.
Review and validate
Check that a request or a document meets what it has to meet.
Classify
Sort what comes in and send it to the right person.
Report
Generate reports and summaries from the data.
Move information
Pass data between systems without anyone copying it by hand.
Examples of what we have already done
Customer support
An agent answers frequent questions by email or web and hands over to a person what it cannot resolve.
Request handling
An agent reads an incoming request, extracts its data and records it in the CRM.
Order tracking
An agent checks the status of an order in the internal system and notifies the customer.
Email routing
An agent reads the emails in a shared inbox and assigns them to the right team.
File review
An agent checks that a case file has all the requested documents and flags what is missing.
Reconciliation
An agent compares orders, delivery notes and invoices and points out the differences for a person to review.
When it helps
When there is repetitive service, review or classification work that people do today, checks that get skipped when there is a rush, or night tasks that could be automated.
Where a person is still needed
- An agent can make mistakes. Decisions with important consequences need review by a person.
- It works with the rules and data sources it is given. Whatever falls outside them should be escalated, not improvised.
- It can only act on the systems it has access to. Without integration, it stays a conversation.
- An agent's error is not just a wrong answer: it can be a wrong action in a real system. That is why its permissions are limited and what it does is logged.
- The texts it processes, such as an email or a document, can contain malicious instructions. An agent should not have more permissions than it needs.
- If the task does not vary, classic automation is cheaper and more predictable.
FREQUENTLY ASKED QUESTIONS
What is an AI agent?
A program that uses a language model to understand a request, decide which steps to take and carry them out with the tools it has access to.
How is an AI agent different from a chatbot?
A chatbot answers. An agent also acts: it queries systems, classifies requests, validates data, generates reports or moves information between tools.
Does an AI agent work without supervision?
It can carry out tasks on its own, but it can make mistakes. Decisions with important consequences need review by a person.
When is an AI agent not worth it?
When the task does not vary and rule-based automation solves it, when you cannot define what it means for the task to be finished, or when a wrong step is irreversible and there is no way to confirm it beforehand.
Can you start with a small agent?
Yes, and it is the recommended way. You start with a narrow use case and human approval on important actions, and extend the autonomy as the results justify it.
What is needed to set up an AI agent?
A specific process, access to the systems involved, the information sources the agent should use, and a clear definition of what it can do on its own and what a person must approve.
Which AI models do you use and where do they run?
We use several models depending on the task: commercial ones, such as GPT or Claude, and others specialized in specific tasks, such as Qwen or Mistral. They run in the providers' cloud or on our own infrastructure that we set up, depending on what the client asks for and needs.
Do people need to be told they are talking to an AI?
Yes. The EU AI Act requires that people know they are interacting with an AI system, unless it is obvious from the context. This obligation applies from August 2026.
What does the cost of an AI agent depend on?
On the number of systems it has to connect to, the volume of interactions, the degree of autonomy and the controls the case requires. Unlike classic automation, each decision the agent makes has a model usage cost.
From the process to the agent
We start with a proof of concept on real data and only scale if the value is demonstrated.