AI glossary · Agents
AI agent
In one sentence
An AI agent is a system that uses a language model to understand a request, decide which steps to take and carry them out with real tools, such as your calendar or CRM, until the task is done rather than just replying.
Also known as: artificial intelligence agent · intelligent agent · autonomous agent
What is an AI agent?
The key word is act. A language model on its own produces text: you ask, it answers. An agent wraps that model in three more things: a goal, a set of tools and some rules. With those it can chain steps together: work out what the customer wants, check for a free slot, book it, send the confirmation and log it in the CRM.
Put it to work
A voice agent that picks up the phone
From €149/mo + VAT
It answers calls 24/7, books appointments and logs every call in the CRM.
See the voice agentThat’s why agents are sometimes called “digital employees”. They don’t replace your team across the board, but they can take on repetitive work with clear rules: answering the phone, qualifying a lead, handling frequently asked questions.
Related to this term
UTIO Voice
- 500 minutes and a number included
- Natural conversation in 26 languages
- Introduces itself as a virtual assistant
- 14-day free trial, no card needed
From €149/mo + VAT
See the voice agentHow it works
An agent runs in a loop:
- Understand the request, spoken or written, using a language model.
- Decide what to do: reply, look something up or use a tool. Your instructions set what it may and may not do.
- Act through integrations: calendar, CRM, database, email.
- Check the result and carry on until the task is closed or handed to a person.
To answer with your own business information, an agent typically relies on RAG, searching your documentation before it replies. And everything it reads and writes is counted in tokens, which is why it should only be given the context it needs.
Example
A hair salon in Dublin misses calls whenever the whole team is busy with clients. A voice agent picks up every call: it asks which service the caller wants, checks the real free slots in the calendar, offers two times and books the one they choose. It then sends a reminder and logs the appointment. If someone asks for something out of the ordinary, such as a quote for wedding hair, the agent takes their details and alerts the team.
AI agents at UTIO
UTIO sets up AI agents for voice, WhatsApp and email with the phone line included: agent, numbers, minutes and CRM in one monthly fee. The voice agent answers 24/7, speaks 26 languages, introduces itself as a virtual assistant and takes actions such as booking appointments in Google Calendar or qualifying a lead.
Autonomy is set in steps: no AI, ask first, draft or automatic. Most businesses start in draft mode and widen the agent’s room from there. You can build the agent yourself from your website in a minute, or ask us to connect it to your calendar and CRM. The tasks agents already run in production are listed under verticals.
Frequently asked questions
What's the difference between a chatbot and an AI agent?
A chatbot talks: it answers questions, usually from a script or from what the model knows. An agent also acts: it checks your calendar, books, updates the CRM or opens a ticket. The difference lies in the tools it can use and in the fact that it works towards finishing a task.
Do AI agents work without supervision?
That depends on how you set them up. The sensible approach is to start with little autonomy, with the agent drafting and a person approving. Once its answers prove reliable, you give it more room, one task at a time.
Should customers be told they're talking to an AI?
Yes, it's good practice and in line with the transparency requirements of the EU AI Act. A well-built agent introduces itself as a virtual assistant at the start of the conversation.
Related terms
- RAG
RAG (retrieval-augmented generation) is a technique in which a language model first retrieves the relevant passages from your own documents and then writes its answer from them, instead of relying on what it memorised in training.
- Token
A token is the chunk of text a language model works with: a word, part of a word or a punctuation mark. Models read, count and bill in tokens, not in words or characters.
Updated on