Client
Fidelity Institutional
Project timeline
2024 - 2 months
Platform
Web
AI Code Generator and Chatbot
Instantly generate new integrations and troubleshoot code
The current API integration deployment process takes 4-6 weeks and involves multiple steps for a developer to retrieve and integrate JSON code for API usage. This tool helps speed up the process.
Problem
The API deployment process, which takes 4–6 weeks, exceeds industry standards. Upon a client’s request, a Fidelity Integration Specialist analyzes the use case, develops a sample API, and engages in continuous testing and iterations, often resulting in significant back-and-forth before it’s ready for integration.
Solution
Offer a self-service AI tool that instantly generates JSON code and includes a chat feature for real-time error troubleshooting. Developers can ideate, test, run sample API calls, review responses, and refine their own APIs before launch.
Impact: Reduce the API deployment process from 4-6 weeks to just 1-3 days for developers
API CODE SAMPLE GENERATOR
CHATBOT
My role
My role was to design and present a prototype to senior leadership, showcasing how we plan to integrate AI into the Integration Xchange platform. This is what I contributed to, as the UX Designer on the team:
1. Wrote the use case
2. Created conversation flows and dialogues
3. Designed the UI template
4. Created wireframes and prototype
5. Presented designs to senior stakeholders
The approach behind my work
AI was a significant focus for our business and product area, with many units exploring how it could enhance their operations. I was invited to design this product due to my experience participating in two immersive AI hackathons at Fidelity—one of which my team won first place. These 3 elements were critical to how I achieved my work:
Participation in AI hackathons & immersive training
Working daily with the Data Science, Data Architecture and Integrations business team.
Soliciting expertise from a Conversation Designer
Learning the API deployment process
For this project, I needed to understand the details of the current API deployment process, so I mapped out the flow (through work sessions with the Integration Deployment team) for reference before starting the design.
Quick switch to plan B: Automation instead of LLM AI
We have a tight deadline to deliver a POC. Initially, we planned to use the ChatGPT LLM model family. However, concerns over its information accuracy and the data team needing more time to work through the technology led us to reconsider. Instead, the team focused on non-AI automation with pre-built use cases.
Global intent patterns
With the switch to automated use cases for this project, I sought expertise from a Conversational Designer in key areas such as interaction patterns, bot humanization, and writing styles. These patterns now guide me in crafting effective user responses.
Utilizing the Cognitive Computing Platform (CCP)
This design is now automated with a chat tool featuring pre-built answers. Instead of waiting for the Deployment team to provide use case examples, I proactively pitched three use cases to keep the process moving forward.
Brand identity & design template
With the scripts reviewed and business feedback captured, the next step was to establish a brand identity and create a template for future designs. This resulted in two templates: a larger version for the code generator and a standard size for the chatbot.
Enterprise-grade API automation solution
The Integration Xchange platform leverages automation to pull code from SwaggerHub, streamlining API integration and development while AI capabilities are still under implementation.
