Agents – Introduction
On the Evoya platform, every interaction with AI runs through an agent. Whether you work in the chat, deploy a bot on a website, automate emails, or process audio recordings – an AI agent is always at the center. This is the core principle of Evoya AI: the agent is the connecting element between you, the AI, and the various channels.
An agent is a configurable AI assistant based on a language model and set up for a specific purpose. It combines an AI model with clear instructions, additional knowledge, and optional capabilities to handle tasks in a targeted way.
How agents and channels work together
An agent is independent of the channel: you configure it once and then deploy it where you need it. In the Chat you work directly with an agent, via Bots you make it accessible on a website, via Email Automation it can process incoming messages, and from Audio Recordings – such as meetings or dictations – it can generate reports or summaries. The agent remains the same element – only the channel through which it is reached changes.
What an agent can do
Agents range from simple helpers to digital employees that take on tasks independently. Even basic agents provide knowledge access to documents, answer questions, and summarize content. More complex agents go well beyond that:
- Automating entire workflows
- Reading and describing data from ERP or CRM systems
- Extracting and structuring information from documents
- Supporting the processing and creation of reports
- Independently taking on tasks as digital employees
How an agent behaves and what it can achieve depends entirely on its configuration.
What agents are used for
Agents are deployed wherever recurring tasks, information queries, or structured interactions need to be automated or supported: in customer service, as internal assistants for teams, for knowledge access to documents, in analysis and research tasks, or in multi-step AI workflows.
Structure of an agent
An agent consists of several components that together determine its behavior. The AI model is the agent's "brain" and determines how it understands, processes, and generates content. The prompts describe the agent's role, tasks, style, and limits. Through data sources, an agent can access additional knowledge – for example from documents or websites. Tools extend an agent's possibilities, such as through searching, reading websites, or analyzing images. Skills are collections of capabilities that can be assigned to an agent – for example, instructions on how to handle Word, Excel, or PDF files. And via subagents, agents can collaborate with other agents and divide tasks among themselves.
The details on the structure and configuration of an agent are covered on the separate subpages of this section.
How to create an agent
The easiest way to start is with the Agent Builder – it guides you step by step through creation. If you want full control, you can also make all settings manually.