Approvals

Approvals allow organizations to introduce human decision-making into AI workflows.

Instead of allowing AI systems to execute every action automatically, Karyam can pause execution and request approval before continuing.

This ensures sensitive operations remain under human control while still benefiting from AI automation.

Why Approvals Exist

Not every AI-generated action should be executed automatically.

Some business processes require human validation due to operational, financial, legal, or security considerations.

Examples include:

Sending high-value payments
Deploying production infrastructure
Approving leave requests
Publishing customer communications
Deleting business data
Executing administrative operations

Approvals ensure that AI assists with decision-making without replacing human accountability.

How Approvals Work

When an AI system reaches a step requiring authorization, execution pauses until a decision is made.

User Request
      ↓
Agent / AI Flow
      ↓
Approval Required
      ↓
Approve or Reject
      ↓
Continue or Stop Execution

Once a decision is recorded, the workflow proceeds accordingly.

Approval Workflow

A typical approval process follows this sequence.

Execution Starts
      ↓
Approval Requested
      ↓
Pending Review
      ↓
Approved / Rejected
      ↓
Execution Continues or Stops

This allows AI systems to safely participate in business-critical processes.

Common Use Cases

Approvals are commonly used for:

Financial Operations

Expense approvals
Purchase requests
Invoice processing
Payment authorization

IT Operations

Infrastructure changes
Server access
Production deployments
Security policy updates

Human Resources

Leave approvals
Employee onboarding
Role changes
Document verification

Business Operations

Contract reviews
Customer communications
Data exports
Compliance workflows

Approval Requests

Each approval request provides the reviewer with the information needed to make a decision.

Typical information includes:

Approval title
Request description
Requested action
Request status
Requestor
Related Run
Timestamp

This provides clear context before approving or rejecting an operation.

Approval Decisions

An approval request typically results in one of the following outcomes.

Status	Description
Pending	Waiting for human review
Approved	Execution is allowed to continue
Rejected	Execution stops
Cancelled	Approval request was cancelled

These decisions become part of the execution history.

Approvals and Runs

Approvals are fully integrated with Runs.

When a workflow reaches an approval step:

Run
    ↓
Approval Requested
    ↓
Execution Paused
    ↓
Decision Recorded
    ↓
Run Continues

This allows teams to understand exactly when human intervention occurred during execution.

Approvals and Audit Logs

Every approval action is recorded within the platform.

Examples include:

Approval requested
Approval granted
Approval rejected
User who made the decision
Decision timestamp

This provides a complete history for operational reviews and compliance requirements.

Governance Benefits

Approvals help organizations:

Reduce operational risk
Maintain human oversight
Enforce business policies
Protect critical systems
Improve regulatory compliance
Build trust in AI automation

Rather than replacing people, approvals ensure AI works alongside them.

Relationship to Other Concepts

Agent / AI Flow
        ↓
Approval
        ↓
Run
        ↓
Audit Log

Approvals introduce controlled decision points into AI execution while maintaining complete visibility.

The Karyam Philosophy

AI should automate repetitive work, not remove accountability.

Approvals combine AI efficiency with human judgment, allowing organizations to automate confidently while retaining control over critical decisions.