2023
Agentic AI - Assistant, 2024
Role
Lead Product Designer
Type
AI Agent
OVERVIEW → 01
Using a live bank API, we allowed our user to query and ask questions on live transaction data, building graphs, writing Python and searching the web with watsonx. This demo is an extension of our work on agents. This was also an exploration into how to understand best practices for designing for agentic workflows and allowing customers to navigate between multiple assets that might have been created throughout the dialogue.
THE CHALLENGE → 02
In the financial sector, users often struggle to extract meaningful insights from their transaction data in real time. While many banking platforms provide static summaries, they lack live interactivity, personalised exploration and agentic workflows that allow users to query, analyse and act on their financial data dynamically. The key challenges were: Real-time transaction data remains largely untapped, preventing customers from gaining actionable insights that could enhance their financial awareness and decision-making. Banking tools stay rigid and generic, failing to adapt to individual user needs or provide personalised recommendations. The technical complexity of building a cohesive system that combines APIs, visualisations and intelligent workflows makes it difficult to deliver a truly integrated experience. While customers interact with various financial data points and assets, they lack a seamless, AI-powered platform that can help them organise information, track tasks and maintain context across different sessions.
THE SOLUTION → 03
This project aims to bridge the gap between real-time data, intelligent agents and user-driven financial exploration. By integrating Langchain, Langgraph, React, and watsonx, we provided an interactive, multi-agent system that enables customers to: Query live transaction data with natural language and in a conversation window, generate tailored insights, visualisations and Python codeSearch the web dynamically to contextualise financial decisions. And finally, seamlessly manage and retrieve previously generated assets.
This exploration not only demonstrates the feasibility of deploying multi-agent financial solutions but also establishes best practices for designing agentic workflows that enhance customer autonomy and empower creativity.
