Agibank connects customers to the right AI agent with Agentforce

Every question, a new conversation

Agibank is a Brazilian digital bank focused on expanding access to financial services for one of the country's fastest-growing populations.

On WhatsApp, one of Agibank’s primary customer service channels, specialized AI agents were already helping customers across a range of needs. But without a way to coordinate multiple agents in one conversation, changing topics meant starting a new conversation from scratch.

Meanwhile, service reps were already handling more than 350,000 calls every month, making it even more important to simplify every interaction while preserving Agibank’s signature personal service.

As the bank set out to grow their customer base by 40% without increasing fixed costs, they saw an opportunity to make every customer conversation faster, simpler, and more connected.

Agent-to-agent orchestration

To make every customer interaction feel like one continuous conversation, Agibank is evolving their AI-powered service experience with a superagent built on Agentforce. While specialized AI agents already support customers across a range of needs, the superagent adds a new orchestration layer that connects these capabilities into a more seamless, unified journey. This helps Agibank scale service while making customer interactions even simpler, more connected, and more personalized.

The superagent receives each WhatsApp message in Portuguese, interprets the customer's intent, and determines which of six Agentforce-powered AI agents should handle the conversation. Whether the customer needs help with a loan, a frequently asked question, retention, or another service, the superagent seamlessly adds the right agent to the chat.

For example, a customer might start by checking the status of a loan proposal, ask a general question about opening a digital checking account, and then request loan portability — all in the same WhatsApp conversation.

As Agibank migrates the six agents and 11 additional bots into a unified agentic architecture, those specialist agents work together under one orchestration layer instead of operating in isolation.

Routing with intelligence

After verifying a customer's identity with their Brazilian identification number against records in Agentforce Service, the superagent then retrieves their account details, active banking products, and open opportunities.

It then uses Atlas Reasoning Engine alongside Agibank's business rules to evaluate the customer’s request against a backend route table, where predefined, deterministic paths lead to specialized AI agents or sales and support reps.

The superagent then connects the inquiry with the right specialized agent:

• Loan status agent: Checks loan proposal status, identifies upsell and cross-sell opportunities, and hands off prospects.
• FAQ agent: Answers routine questions about products, processes, and services.
• SDR agent: Engages warm leads from marketing campaigns, answers questions, and qualifies sales-ready opportunities.
• Retention agent: Guides customers through portability requests, explains contract terms and requirements, and completes retention.
• PSAT agent: Sends customers a satisfaction survey after a service interaction.
• Autoservices agent: Handles banking and financial inquiries for customers across credit cards, loans, and account services.

If a request could follow more than one path, the superagent presents a clear, numbered list of options for the customer to choose from. Once a route is selected, the full context stays with the interaction so customers can switch topics without starting over.

To keep interactions secure, the superagent is limited to specific topics and automatically routes conversations involving compliance or fraud concerns to a human rep when business rules require it.