BrazilFinance and insurance2025 – ongoingAgibank
Salesforce Agentforce AI agents for WhatsApp customer service
Agibank's AI agents on WhatsApp deflected 80% of loan-proposal queries, per Salesforce
Problem
Agibank serves customers with monthly incomes up to R$ 3,5 mil, and most of its service runs on WhatsApp. Its older chatbot followed a decision tree.
Approach
Agibank replaced the decision-tree bot with several AI agents on Salesforce's Agentforce platform, linked to its core banking system. A routing agent sorts the request. A FAQ agent answers from the knowledge base, and a loan portability agent handles customers moving loans elsewhere. Human agents keep the complex cases.
Results
Takeaway
Splitting one bot into narrow agents let the bank measure results by use case. The loan-status agent and the FAQ agent showed very different deflection rates.
Categories
- Industry
- Finance and insurance
- Business function
- Customer service
- Innovation type
- Process innovation
- AI technology
- AI agentsGenerative AI
- Adoption
- Off-the-shelf tool or platform
- Evidence
- Vendor case study
Sources
Source summaries
Salesforce's press release describes Agibank's multi-agent setup on WhatsApp, which replaced a decision-tree chatbot. It gives deflection rates of 80% for credit-proposal queries and 30% for FAQs, and 70 mil queries a month for the loan-status agent. It also lists a target to double the number of agents in production, which is not a result.
Full details are in the original sources.