Automation & Productivity

Turning repetitive work into automated pipelines, so your team focuses on what actually matters.

Chatbot CHR

The Challenge

Every missed booking outside business hours is a permanently lost customer. In the CHR sector, 60 to 70% of reservation requests arrive in the evening or on weekends, when no one is available to respond.

Staff was losing precious time answering the same questions over and over: availability, pricing, menus, location. Hours spent on WhatsApp and Messenger instead of serving the clients already there.

The Solution

I built an AI chatbot integrated into the client's messaging channels:

  • AI FAQ with Memory: The chatbot answers all recurring questions instantly: pricing, availability, menus, hours, location. Available in French, English, or Malagasy, 24/7.
  • Booking Capture: Automatically collects name, date, party size and saves directly to Google Sheets. Zero manual entry.
  • RAG Knowledge Base: The chatbot is trained on the property's real data (rooms, special offers, amenities) and responds accurately, not with generic answers.
  • Smart Escalation: If the request exceeds its capabilities, it transfers to a human agent with the full conversation history.
  • Cross-session Memory: The chatbot recognizes returning customers and adapts its responses based on their history.

The Impact

24/7 Availability
-80% Manual Response Load
+40% After-Hours Bookings

Gallery

πŸ’‘

Not in the CHR sector?

This demo was built for a hotel/restaurant, but the architecture is fully reusable. The chatbot can be adapted to any industry: real estate, healthcare, e-commerce, professional services. With a knowledge base built around your specific business.

Discuss an adaptation

OCR Document Processing

The Challenge

An accountant or admin manager can spend 8 to 12 hours a week manually re-entering data from paper invoices, delivery notes, and scanned receipts. A slow, repetitive process and a source of costly errors.

The goal: zero manual entry, zero silent errors, and a pipeline that runs even overnight.

The Solution

I built an OCR document automation pipeline triggered directly from WhatsApp:

  • WhatsApp Trigger: The user simply sends a photo of their invoice via WhatsApp. That's all they need to do.
  • AI Extraction: Gemini Vision analyzes the image and automatically extracts key fields: date, pre-tax/tax amounts, invoice number, supplier name, VAT.
  • Validation and Cleanup: Extracted data is normalized and validated before writing. No misformatted amounts, no ambiguous dates.
  • Structured Output: Data lands directly in Google Sheets, ready to use. No copy-paste.
  • Exception Flagging: If extraction confidence is low (blurry image, atypical document), the document is flagged for human review. No errors pass silently.

The Impact

Γ—10 Faster Processing
-95% Entry Errors
8h Saved per Week

Gallery