Flows are the core building blocks of every AI agent in Thena. Each flow defines a specific task the agent can perform—from generating titles to deflecting tickets—with full control, customization, and observability.

What are flows?

  • Each AI agent can have multiple flows, representing different capabilities
  • Every flow is defined by a prompt — the instructions that guide how the agent performs the task
  • Flows are tied to workflows — you decide when and where they run using triggers and filters
  • They are powered by native LLMs with support for additional knowledge via uploaded files or external links
  • All flows use the agent’s tone, language, and personality settings to ensure brand alignment

Prompt-driven intelligence

Every flow has a prompt that acts as its instruction manual. You can customize these prompts to:

  • Define the agent’s behavior
  • Control the structure and tone of the output
  • Add constraints or formatting rules

This allows teams to fine-tune flows for different products, teams, or customer segments.

Knowledge + LLM

Flows are powered by a built-in large language model (LLM), but that’s just the start. You can enhance flows by uploading:

  • Files (PDF, TXT, CSV, DOCX)
  • URLs (documentation, internal help, knowledge base)

This gives every flow access to the right information, making its output more accurate and contextual.

Learn more about how to enhance your agents with domain-specific information in the Knowledge guide.

Workflow-based execution

Flows are not always-on. You control when and where a flow is triggered using Thena Workflows. This gives you:

  • Granular control based on fields like priority, channel, tags, sentiment, assignee, and more
  • The ability to scope flows to specific accounts or segments (e.g., “only for commercial tickets on email”)
  • Smart automation that blends AI with your business logic

Execution logs & observability

Every time a flow is triggered, Thena logs:

  • When it ran
  • What triggered it
  • Which input it used
  • What output it generated

You can track these executions both from the Agent panel and the Workflow execution view.

This gives you full visibility into how your agents operate—and confidence that your automation is working as expected.

For a detailed view of execution history and analytics, check out the Executions guide.