Agentic AI Workflows
Learn how to build, orchestrate, and deploy intelligent multi-agent workflows inside BizFirstAI Studio — from data ingestion and AI classification through to Elasticsearch persistence and automated notifications.
What is an Agentic Workflow?
An agentic workflow is a process where AI agents — equipped with tools, memory, and reasoning — autonomously execute multi-step tasks, make decisions, and take actions to achieve a goal without step-by-step human instruction.
In BizFirstAI Studio, each agent is a node on a visual canvas. Agents expose typed ports that define how data flows at runtime and how services are composed or inherited at design time. Before execution, BizFirstAI resolves all static relationships (inheritance, composition) into a unified agent configuration.
Three Core Concepts
This guide walks through three progressively deeper demonstrations of agentic workflows in BizFirstAI Studio.
Fetch & Insert Workflow
An end-to-end pipeline: fetch order data from a vendor API, transform it with jq, classify it with an Octopus AI Agent, insert it into Elasticsearch, then send a Gmail notification.
AI Agent Ports
Understand the six port types on an AI Agent node — Input, Output, Component, Sub Agent, It Extends, and My Components — and how runtime vs. static composite relationships differ.
Deep AI Agent
Design hierarchical multi-agent topologies: a shared base agent, an orchestrating main agent, and multiple participant sub-agents — all resolved into a unified configuration before runtime.
Key Design Principles
Canvas-Based Workflow Design
Every node, connection, and agent is placed on a visual canvas in BizFirstAI Studio. Workflows are composed by dragging nodes and connecting ports — no code required for orchestration logic.
Strict Port Contracts
Ports define the contract between nodes. Runtime ports (Input/Output) carry data at execution time. Static ports (Component, It Extends, Sub Agent) define structural relationships that are resolved before execution.
Design-Time Resolution
All inheritance and composition chains are resolved at design time into a single consolidated configuration per agent. This means complex multi-agent hierarchies execute with the simplicity of a single agent at runtime.