The Client’s Challenge
Leveraging its extensive API maturity and pioneering use of WhatsApp bots for transactions, the institution identified an opportunity to become a true agentic connectivity hub. One of the project’s main drivers was to provide an intelligent assistant to support the lending journey by qualifying leads and reducing the cost of unnecessary analyses before applications reached the approval pipeline.
Sensedia’s Role
The architecture incorporated Sensedia AI Gateway as an enabler from the outset, connecting the AI agent to the bank’s lending services via MCP (Model Context Protocol). It delivered centralized AI governance (cost-per-token monitoring, observability, rate limiting, and prompt guardrails), domain control (prompt enrichment and moderation to ensure the agent operated strictly within business rules), and secure integration (a clearly defined scope within a protected environment, safeguarding corporate systems).
Results
In just two weeks, the institution achieved milestones that demonstrated the model’s value: end-to-end validation (the complete flow - including acceptance of terms, Brazilian taxpayer ID [CPF] and age validation, installment rules, and lead submission to the operational pipeline - ran successfully in a real-world environment); value generation (real credit leads were captured and qualified); proven security (the AI agent rejected queries outside the lending scope); and increased maturity of the initiative (the AI assistant moved into production, consolidating the agent banking model).
.webp)
.svg){kind=logo}