Applied AI experiment
HR AI Agent
Leave requests through LINE and a review portal.
An experiment connecting a LINE conversation with n8n, local language-model processing, structured leave data, and a web interface for review.
Project context
My contribution
I built the prototype flow across webhooks, intent and detail extraction, conversation state, database updates, and the review interface.
Scope
The experiment explores conversational leave intake while keeping approval and rejection decisions with a human reviewer.
System view
What the work focused on
ExperimentEmployee flow
Conversation
Collects leave type, dates, and context through familiar chat interactions.
Automation
Orchestration
Connects messaging, local model processing, and stored data through n8n.
Decision point
Human review
Leaves the approval or rejection decision in a web review interface.
Delivery
What the work required
- Connected LINE Messaging API with n8n to receive and respond to employee messages.
- Prompted for missing leave details and converted conversational Thai into structured data with a local model.
- Built a web review interface for request status updates and return messages through LINE.
Prototype flow
Conversation and review
A conversational intake and web review flow for converting messages into structured leave requests.
- Asks follow-up questions when required information is missing.
- Creates a structured request for a reviewer rather than making the decision automatically.
- Returns status changes through the messaging workflow.
Conversational intake
A leave request converted into structured information.
Request review queue
A review view with extracted request data and status controls.