# BRIX AI

> [HTML Version](cortex.html)

[BRIX AI](https://brix365.com/en/products/ai-cortex/) is a solution designed for end-to-end integration of artificial intelligence (AI) technologies into your enterprise. It leverages large language models to automate employee tasks and accelerate routine operations, enabling you to:

- Set up AI agents that answer employee queries in a chat and execute actions within the system.

- Embed AI operations into corporate business processes, user interfaces, and scripts.

- Enable smart search across corporate knowledge bases.

This solution is available exclusively for BRIX On-Premises deployments. The [BRIX AI](https://brix365.com/en/products/ai-cortex/) server is installed directly within the client's infrastructure.

## BRIX AI terminology

Key concepts related to **BRIX AI** include:

- **Large language model **(**LLM**):** **An artificial intelligence system capable of analyzing and generating text. **BRIX AI** supports ready-to-use language models from various providers, as well as fine-tuned models.

- **Fine-tuning**:** **The process of adapting a pre-trained language model for specific tasks by further training it on a custom dataset.

- **Token**:** **The basic unit of text processed by a language model, such as a word, subword, or character. Model usage provides a specific allocation of tokens, which are consumed when analyzing prompts and generating responses.

- **BRIX AI components**:** **AI-driven product capabilities, including:

	- **AI agents**:** **Virtual assistant bots that interact with users in chat interfaces across BRIX. They answer natural language queries, process data, and execute actions within the system.

For example, an employee can ask an AI agent to find all contracts from last month, sort them by total value, and assign follow-up tasks. Future releases will introduce autonomous operation, enabling AI agents to run in the background without user intervention.

	- **AI operations**:** **Specialized, single-purpose AI agents configured to perform a specific function or task. They deliver predictable outcomes by using predefined input and output formats, such as text or JSON objects. An AI operation can be triggered within a business process, executed from a form or page, or called via API from a script.

For example, during customer email processing, an AI operation can generate a brief summary and display it directly within a user's task.

	- **Smart search**:** **Vector (semantic) search technology that delivers answers based on query intent and context. Supported data sources include websites, documents, Confluence, and data stored in BRIX app items.

Smart search projects connect to AI agents and AI operations to supply them with precise data retrieved from connected sources. For example, employees can ask an AI agent questions regarding internal company policies and regulatory documentation.

- **Tools**:** **Functions invoked by agents and AI operations to perform calculations, trigger actions, retrieve information, and more. A given agent or AI operation's capabilities depend on its connected tools.

In **BRIX AI**, you can use built-in tools, configure functions to interact with BRIX apps and business processes, or extend capabilities by connecting custom tools via MCP servers—for example, to pull data from external systems.

## BRIX AI product architecture

The product consists of two main components:

1. The **BRIX AI** server: Installed on-premises within your infrastructure. It features a management designer that includes the following sections:

- **Agents**.

- **AI Operations**.

- **Smart Search**.

- **BRIX Tools**: Used to configure custom tools that extend AI agent and AI operation integrations with BRIX.

- **Providers**: Used to connect large language model providers.

- **API Keys**: Used to configure integration with BRIX or third-party systems.

2. The **BRIX AI Integration** solution: Provided upon request, and includes a module of the same name that delivers:

- Integration configuration between BRIX and the **BRIX AI** server.

- User access management for AI agents.

- AI agent chat widget, automatically displayed on user pages for accounts granted access to at least one AI agent.

- The **Trigger AI Operation** widget, containing an AI operation selection field and action button. This widget can be placed on any form or system page, allowing users to execute pre-configured **BRIX AI** operations at the click of a button.

- The **Trigger AI Operation** process element, which can be added to business processes to run an AI operation at a designated workflow step.

## BRIX AI editions

Depending on your business scale and AI deployment scenarios, you can choose from the following available editions:

- **BRIX AI Lite**

- **BRIX AI Standard**

- **BRIX AI Enterprise**

To learn more, see [Available BRIX AI editions](cortex-editions.md).

## BRIX AI licensing

To use the product, you need to:

1. Select a suitable edition and acquire a [server license](licences.md#server) for** BRIX AI**. Key features of this license model include:

- Fixed cost independent of token usage volume.

- Ability to configure unlimited components included within your edition.

- Full feature availability for all licensed BRIX users.

2. Provide access to large language models. You can connect cloud services or deploy self-hosted models within your infrastructure. Model usage costs are paid directly to provider vendors and are not included in the software license.

3. Optionally connect external analytics services, such as the Arize Phoenix platform for tracing AI agents, AI operations, and smart search performance.