Case Study
Number AI Assistant
Number provides powerful and flexible payment solutions, supported by extensive technical documentation for developers. However, navigating this vast library of information could be time-consuming, creating friction for developers during the integration process. As a proactive initiative, we designed and built the Number AI Assistant—a conversational AI embedded directly within their docs. This tool gives developers instant, accurate answers and code snippets, dramatically speeding up integration cycles and simplifying complex technical challenges.
Technology
2025
Services
Ideation workshops
UX/UI Design
Gen AI Integration
Quality Assurance
Prompt Engineering
Technology Used
GPT 4o
Azure AI Search
React
Python 3.11
Flask

Client feedback
Challenges

Ensuring 100% Factual Accuracy
For a payments provider, providing incorrect integration advice is not an option. The primary challenge was to eliminate the risk of AI "hallucinations" and guarantee every answer was reliable, verifiable, and sourced directly from the official documentation.
Customizing a Pre-Built Solution
To meet the rapid timeline, we utilized a pre-built Azure RAG tool. However, its default UI was generic and clashed with Number's modern brand identity. The challenge was to completely redesign the assistant to create a seamless, on-brand user experience.
Accelerating, Not Just Answering
A simple Q&A bot isn't enough to significantly speed up development. The assistant needed to go beyond text answers and actively accelerate the integration process by providing ready-to-use code snippets and clear pathways to deeper information.


Solutions

Robust Prompt Engineering & Source Linking
We implemented a rigorous prompt engineering strategy to constrain the AI's responses strictly to the source documentation. Crucially, every answer generated by the assistant includes linked citations, allowing developers to instantly click through to the source page to verify the information and build essential trust.
Custom Front-End Design & Styling
Our designer and developer worked in tandem to apply a custom front-end layer over the Azure tool. We meticulously implemented Number's design system—including typography, colors, and interface elements—to make the AI assistant feel like a native, fully integrated part of their developer portal.
Actionable Code Snippets & Seamless Navigation
The AI was trained to identify when a code example is needed and can paste formatted code snippets directly into the chat. Combined with source links that open in new browser tabs, this allows developers to get what they need and dive deeper without losing their place, dramatically improving the developer experience.

Results
Hire usSignificantly Reduced Developer Onboarding Time
The Number AI Assistant enabled developers to quickly access accurate and context-specific integration details, eliminating the need to manually search through large volumes of technical documentation. By providing instant, source-linked responses, the assistant allowed new team members to understand available payment methods for each integration type within minutes rather than hours or days.
Improved Accuracy and Consistency in Payment Method Assignments
Before the AI Assistant, developers often spent significant time cross-referencing various documentation sources to determine which payment methods were compatible with specific integration channels. This process was prone to misinterpretation and inconsistent implementation across teams. This improved consistency reduced the risk of integration errors, prevented costly rework, and ensured compliance with security and data-handling standards.

