Get Started with AI for Process

Set up a workspace, configure AI models, and build your first AI-powered workflow in AI for Process.

Prerequisites

• Kore.ai account with AI for Process access.
• Admin permissions for initial setup.
• (Optional) API credentials for external integrations.

Access AI for Process

AI for Process is currently available on our US Cloud at https://process.kore.ai. To request access for your customers, partners, or prospects, contact the Sales team.

1. Navigate to the Platform.
2. Select AI for Process from the product menu.
3. Choose or create a workspace.

Workspace Setup

Create a Workspace

A workspace is a logical grouping of workflows and configurations, typically organized by teams or projects.

• A default workspace is created automatically for each account.
• The current workspace appears in the application header and can be changed by using the workspace switcher.
• Switching workspaces updates the workflows, settings, and permissions in the interface.

Invite Team Members

1. Go to User Management.
2. Click Add new users> Invite.
3. Assign roles:
   • Admin — Full workspace access.
   • Developer — Create and edit workflows.
   • Reviewer — Process Inbox tasks.
   • Viewer — Read-only access.

Workflow Setup

Set up your AI-powered Workflows in three steps:

┌──────────────────────────────┐ │ Step 1: Configure Models │ │ Set up AI model providers │ └──────────────┬───────────────┘ │ ▼ │ ┌──────────────────────────────┐ │ Step 2: Build a Workflow │ │ Create an AI-powered │ │ workflow (Add nodes,trigger,│ │ human review) │ └──────────────┬───────────────┘ │ ▼ │ ┌──────────────────────────────┐ │ Step 3: Test and Deploy │ │ Test your workflow and │ │ deploy it for use │ └──────────────────────────────┘

Configure Models

View Available Models

1. Navigate to Models.

2. View available models:

   • Open-source — 30+ popular models, or any text generation model from Hugging Face.
   • External — Commercial models including OpenAI, Anthropic, Azure OpenAI, Cohere, and Google. You can also integrate your own hosted models via API connections.
   • Fine-tuned — Models fine-tuned for your specific use cases.

Add an External Model

1. Go to External Models.
2. Click Add Provider.
3. Configure credentials:

For example:

Provider: OpenAI
API Key: sk-...
Models:
  - gpt-4
  - gpt-4o
  - gpt-3.5-turbo

Use the Model in a Workflow

Select the configured model when setting up nodes in the workflow builder.

For example:

AI Node Configuration:
  model_source: external
  provider: openai
  model: gpt-4
  temperature: 0.7
  max_tokens: 500

Build a Workflow

Step 1: Create a Workflow

1. Navigate to Workflows.
2. Click Create Workflow.
3. Enter a name and description for your workflow.

For example:

• Name: “Document Summarizer”
• Description: “Summarizes uploaded documents”

Step 2: Add Nodes

Build your workflow by connecting nodes in the flow builder.

The following example shows a simple AI-powered workflow:

[Start] → [Text-to-Text AI] → [End]

1. Start Node — Added automatically. Configure input and output variables, and optionally set up triggers or a schedule.
2. Text-to-Text AI Node — Add from the node palette.
3. End Node — Add to complete the workflow.

AI for Process supports multiple node types — including AI, Integration, API, Function, Condition, Loop, Browser, and more. Add the nodes that match your workflow logic. For a complete reference of all available node types, see Workflow Nodes.

Step 3: Configure Nodes

For example:

Node: Summarize Document
Type: text-to-text
Model: gpt-4
Prompt: |
  Summarize the following document in 3-5 bullet points:

  {{input.document_text}}

  Focus on key findings and action items.
Output variable: summary

Step 4: Configure Triggers and Schedulers (optional)

The Start Node supports event-based triggers and scheduled execution to automate workflow initiation.

Set how the workflow is triggered:

Trigger type	Description
Event	Starts the workflow when a specific event occurs in a connected third-party app.
Schedule	Starts the workflow at a set time or interval.

For configuration details, see Start Node.

Step 5: Test and Deploy

1. Click Test to run with sample input.
2. Review output and logs in debug panel.
3. Click Deploy to make available.
4. Copy API endpoint for integration.

Add Human Review

If your workflow requires human validation, add a Human Node at any point in the flow.

[Start] → [Text-to-Text AI] → [Human Review] → [End]

When to use it: approval workflows, quality assurance checkpoints, compliance reviews, or exception handling.

For example:

Human Node: Review Summary
Task type: approval
Assignees:
  - role: reviewer
Instructions: |
  Review the AI-generated summary.
  Approve if accurate, or edit and approve.
Fields:
  - name: summary
    type: text
    editable: true
  - name: approved
    type: boolean
Timeout: 24h

For configuration details, see Human Node and Human Review.

Use the Inbox

1. Navigate to Inbox.
2. View pending review tasks.
3. Review AI output.
4. Approve, edit, or reject.
5. Workflow continues after action.

---