Overview¶
Agentic Workflows are the foundation of the OI Agents’ automation system. They enable you to design, coordinate, and execute AI-native workflows using intelligent agents that interact with data, tools, and systems to perform structured, multi-step tasks.
Unlike linear scripts or simple prompt-based flows, agentic workflows support dynamic reasoning, decision-making, and tool invocation, enabling true automation across complex business processes.
What Makes a Workflow "Agentic"?¶
An agentic workflow is built from the following core elements:
- Agents: Reusable logic units powered by LLMs, configured with models, system prompts, tools, and memory. Each agent can reason, retrieve, and act based on context.
- Tools: Functional components (e.g., API calls, document readers, database queries) that agents use to take actions in the workflow.
Key Capabilities¶
- Multi-agent orchestration: Assign different agents to different tasks with collaborative handoffs and conditional logic.
- Tool-augmented reasoning: Equip agents with tools to extend their capabilities beyond generation.
- Enterprise integrations: Connect to external systems using MCP or HTTP tools, with authentication and schema support.
- Reusable components: Use libraries of agents, tools, and templates across workflows to reduce duplication.
- Secure and observable: Built-in role-based access, audit logs, and evaluation hooks for production-grade governance.
Common Use Cases¶
- Automated decision trees (e.g., legal triage, financial approvals)
- Agent-assisted document processing (e.g., contract extraction, policy review)
- Cross-system coordination (e.g., sales ops pulling CRM + ERP + email data)
- Enterprise copilots with background agents executing follow-ups, validations, or escalations