Programs

Introduction to Programs

A Program in cortx is a structured, AI-orchestrated execution flow designed to manage real-world business processes from start to finish. Unlike single workflows or documents, a Program represents a multi-event journey—from data setup to task automation—across datasets, tools, and agents.

Programs are typically initiated through chat, guided by agents like Navi or a designated Program Builder agent. These flows are dynamic, context-aware, and built to automate repetitive operations such as Re-KYC, Customer Onboarding, Claims Verification, and beyond.

Program Building

  • A multi-step, goal-oriented flow designed around a specific business process.
  • Triggered and managed via a conversational agent (e.g., Program Builder).
  • Executed by multiple AI agents, connected datasets, and external tool integrations.
  • Powered by logic-aware prompts, ensuring the right data and actions are handled at each step.

Each Program is tracked end-to-end, with visibility into dataset mappings, task execution, and completion status.

Program Anatomy

  • Program name: The label for your initiative (e.g., “Re-KYC Campaign – Q3”)
  • Goal/Intent: The business objective driving the Program (e.g., “Update KYC for users”)
  • Tool Connection: A linked internal/external system (e.g., My Databricks) where data resides
  • Dataset: Table or collection pulled from the connected tool
  • Schema/Table: Defines the structure of your dataset (e.g., default schema, users_kc)
  • Trigger Events: Logical steps (e.g., send email, create task, mark complete)
  • Agents: cortx agents (e.g., Navi, Analyst) orchestrate Program execution
  • Outputs: Tasks, messages, documents, reports, or synced data

How to create a Program

Step 1: Start the Program Builder

  • Navigate to Home → Start New Chat.
  • Select Program Builder or type: Create a new program.
  • You will be prompted to enter the Program Name.

Step 2: Define the Program Intent

  • The agent will ask what you’d like to achieve.
  • You can specify a goal. For example, “Vendor credentialing for new partners.”

Step 3: Select or Create a Dataset
You’ll be prompted to either use an existing dataset or create a new one.

  • If using existing dataset:
    • The agent will list available datasets.
    • Choose the relevant one (e.g., users_kyc_2025).
  • If creating a new dataset:
    • Select a Tool (e.g., My Databricks) — this acts as the external connection.
    • Choose the Schema (e.g., default) and then the Table (e.g., users_core_kyc).
    • Select required Columns.

Note: Only selected columns are pulled into cortx. Choose wisely based on what agents need to function.


Step 4: Validate and Lock Dataset
Once columns are selected, the dataset is validated and locked for the Program. This ensures clean, consistent execution.

Step 5: Launch Program

  • Agent confirms readiness.
  • Program begins execution: messages are sent, tasks are generated, follow-ups are tracked.

Best Practices

  • Use clear Program names for easy tracking across teams.
  • Include Email and Mobile in datasets for communication-based flows.
  • Test on DEV environments before going live (e.g., My Databricks Z).
  • Assign owners or team roles via the chat agent if applicable.
  • Rely on cortx agents to perform execution, follow-ups, and status monitoring.

Was this article helpful?

on this page