-
-
Client Facing Consultant Interface
-
Backend-Admin Analytics interface
-
Plan matching logic
Inspiration
A friend is an actual pension consultant in the Italian market, this the implementation of an agent able to deliver great investment plan advice in a fraction of a time and 24x7.
What it does
Welcomes and collects personal, financial, employment and other information relevant to the Italian market and it provides a list of investment funds/pension plans depending on the client profile both using hard regulatory filters and logic as well as using Voyage AI to re-rank and label affinity of the plans based on a semantic score. Also a backend Analytics Mongo MCP agent is implemented for consultants of the company to query the MongoDB directly.
How we built it
ADK on Python with Docker Compose, Make orchestration for local and Cloud Run to deploy the agent, front-end with a MongoDB backend. In the background Gemini Flash is used to handle the customer experience and profile information collection. Then and internal matching engine is used to find the right plans and Voyage AI on MongoDB helps with extra relevance and sorting evaluation on top of the empiric one. There is also an internal agency aimed agent that integrates with MongoDB through its MCP implementation, thus allowing personnel from the agency to query the application status or ask about generic information from the database such as Pension specific plans information.
Challenges we ran into
The main challenges were quickly learn a totally new ecosystem and try to implement as much added value as possible. I also wished i could implement a full consultancy experience by integrating real data and real payment checkout solution but had to limit it due to the lack of time and resources.
Accomplishments that we're proud of
Great added value user experience with the possibility to quickly provide relevant information through the agent, using extra AI functionality to provide better matching as well as flexible backend user experience for the office personnel.
What we learned
Great experience learning more python, ADK and Google AI products and frameworks with the MongoDB, Voyage AI and the MCP interface for data/profile and information holding.
What's next for Consulente Pensione AI
Ideally this can be used as a realistic commercial demo for the insurance/pension industry to show how Agents can help provide a more extended and flexible user experience in this complex industry of pension funds where usually the main barrier is to build a relation and connection with the clients in order to provide them with the right advice.
Log in or sign up for Devpost to join the conversation.