Call of Agents

Email response to "I want to check all the accounts I have."
Email response to "How do I open up a bank account?"
AI chatbot on web portal
Receipt processing on web portal
Admin dashboard on web portal
Call history on web portal

Inspiration

Long customer service calls often waste time as users are keep pressing different numbers to request, waiting to be connected to human agents, being transferred between multiple human agents. We wanted to eliminate that frustration and create a single, intelligent system that delivers answers instantly.

What it does

Call of Agents automates service calls using a multi-agent AI system. With just one call, users are authenticated, their intent is analyzed, data is securely retrieved, and responses are delivered—no waiting, no stress.

!!! It's NOT just a Chatbot !!!

How we built it

Multi-agent architecture framework w/ agent supervisor: CrewAI
LLM: NVIDIA Nemotron Nano 9B V2
LLM Provider: OpenRouter
VLM: Gemini 2.5 Flash
Relational DB: PostgreSQL (authentication & RDB agent)
Vector DB: Milvus (governance, RDB and FAQ agent)
7 Worker Agents: Authentication, Intent Analyzer, Governance, RDB, FAQ, Archive, and User Interaction

Supervisor chooses the worker agents DYNAMICALLY (between RDB and FAQ) based on use case, not in the static workflow.

Challenges we ran into

Designing smooth, dynamic communication between multiple agents.
Ensuring governance rules prevent unauthorized data access using RAG.
Maintaining context between agents.
Integrating different external services/protocols (Twilio-phone call, OpenRouter-llm, SMTP-email)
Building & connecting different tools to agents (Postgres query execution, RAG-based SQL query generation, email)

Accomplishments that we're proud of

Fully automated end-to-end customer call process.
Built a scalable, architecture using multi-agent orchestration. (SO EASY TO ADD, PLUG & PLAY WORKER AGENTS!)
Achieved real-time responses with minimum waiting time.
Integrated governance without compromising speed.

What we learned

We learned how to orchestrate complex agent communication, integrate multiple enterprise-level databases, and balance automation with compliance. We also deepened our understanding of AI-driven workflow design.

What's next for Call of Agents

We plan to extend Call of Agents to other industries like insurance, HR and supply-chain, more access points, multilingual voice support, more external service integration, improving customer experience.