Skip to main content

Main Concepts

The main concepts of branchly are centered around providing a structured and intelligent system for managing customer interactions via AI-enabled user interfaces. Below is a concise explanation of these key concepts.

branchly - main concepts

Organization​

In the branchly ecosystem, dashboard users operate within an organization. You have the option to create a new organization or join an existing one. Your subscription and its associated usage limits are linked to your organization, allowing you to create as many Applications as needed within these limits.

Application​

An Application acts as a comprehensive repository for knowledge, logic, and the user interface elements of an AI-enabled use case within an organization. An organization can have many applications.

tip

When does it make sense to have multiple Applications? When you have distinct use cases and the data / user requests should not be mixed but rather focused. An indicator could be that you use different user interfaces at different website locations and you want the user interfaces to behave differently.

User Interface​

branchly offers four distinct user interfaces for embedding into your website or application. Each user interface is customizable in style, placement, and behavior through the dashboard:

  • Chat: A conversational interface (available as a floating Chat Widget or inline page embed) powered by generative AI and custom agents to answer customer inquiries.
  • Search: A search bar interface utilizing semantic-hybrid search, keyword highlighting, and AI-generated answers to guide users to relevant pages.
  • Navigator: An interactive self-service layout combining search and trend-based navigation nodes for structured, click-based website exploration.
  • Form: A smart AI form that adapts fields dynamically in real time, automatically answers routine questions, and routes inquiries to designated channels.
  • API (Headless Interface): A headless interface providing REST API endpoints for forms, chat, and search, allowing you to integrate branchly directly into custom environments (e.g. WhatsApp for Business, Slack channels, or custom mobile applications).
  • Voice & Telephony: Integration options enabled through partners to handle automated voice communication and call center routing.

Knowledge Base​

The knowledge base within branchly is structured into various nodes, each with its type/label, HTML content and title. These nodes are interconnected through relationships to facilitate the dynamic navigation, as users often prefer clicking through related topics. You can manage nodes either individually via the Node Editor or through data sources. The Node Editor includes a WYSIWYG HTML editor for content editing. We use a multilingual design by default, enabling content addition and updates in multiple locales, allowing for language and country-specific customization.

Data Source​

Data sources populate your knowledge base with content nodes. branchly supports a range of sync options:

  • File Upload (CSV, PDF): Processes text or structured data from uploaded documents.
  • Web Crawler: Extracts relevant HTML content from your website automatically.
  • WordPress: Connects to your WordPress site via the REST API v2 to import posts, pages, or custom entities, and uses webhooks for real-time syncing.
  • HelpSpace: Connects a HelpSpace site to sync support articles into your knowledge base.
  • Databases: Connects with PostgreSQL/MySQL databases.
  • OpenAPI Specification: Allows integration of REST APIs by uploading a valid OpenAPI specification in JSON format.

Most data sources can be scheduled for automatic content updates using a cron schedule.

AI Action​

AI Actions (formerly referred to as Tools) represent a powerful concept in branchly. They extend the capability of the branchly generative AI agent from pure information retrieval (RAG) into active process automation.

When using the agent chat strategy, the AI has access to a suite of activated AI Actions (such as document search, Calendly scheduling, or Google Maps directions). Rather than relying on rigid predefined workflow builders, the AI agent dynamically chooses the most appropriate action in real time to resolve a user request, retrieve external information, or execute a workflow.