AI Agents on WhatsApp: From Automation to Business Intelligence

The Transactional Agent: iFood Gerente

iFood Gerente represents the most advanced agent type in terms of operational capabilities: an agent that doesn't just provide information but actually executes financial transactions in real time.

Capabilities: - Pix payments to suppliers and employees - Bill payments with barcode scanning - Real-time balance and statement inquiries - Financial reports and projections

Architecture: - Integration with Zoop (iFood's fintech) for transaction processing - Multi-factor authentication for security - Conversational interface for complex financial operations

The transactional agent represents a paradigm shift: WhatsApp ceases to be a communication channel and becomes a platform for executing business operations. This opens possibilities that go beyond customer service — enterprise process automation, internal operations, and financial management.

The key challenge is security: transactional agents require robust authentication, encryption, and auditing. Zoop's partnership brings the necessary fintech infrastructure, but the principle applies to any transactional implementation.

The Commercial Agent: Lu do Magalu

Magazine Luiza's Lu represents the most complete commercial agent: a multi-agent orchestration system that handles the entire purchase journey from discovery to payment.

Capabilities: - Natural language product search across 37M+ offers - Organized comparisons with detailed specifications - Cart management with coupons and shipping - In-conversation payment (Pix, credit card, installments) - Tracking and after-sales

Architecture: - Multi-agent orchestration (search, comparison, cart, payment, after-sales) - Google Gemini + proprietary models - Integration with product catalog, pricing, and payment systems - Preservation of Lu's personality across all interactions

The commercial agent is the most complex to implement because it requires integration with multiple systems (catalog, pricing, inventory, payment, logistics) and coordination between specialized agents. But the results justify the complexity: 3x conversion vs. the app, NPS 90, and 1M testers.

Lu also demonstrates the importance of personality in commercial agents. The trust built by the character over years translates into higher conversion, proving that in conversational commerce, who sells matters as much as what is sold.