ADB Navigator transformed how ADB's 3,000+ consultants access half a century of institutional knowledge. By designing an AI-powered conversational interface that combines the citation capabilities of Perplexity with intuitive information discovery, we eliminated the "who knows what" bottleneck. The solution features a prompt library for consistently high-quality responses, displays key personnel contacts for direct follow-up, and surfaces related knowledge for comprehensive research.
User satisfaction was immediate, with one senior consultant explicitly praising the intuitive experience. The end result: 40% reduction in poor quality responses and critical information now accessible within five clicks—turning scattered data into actionable insights.
Data Hell
Enterprise challenges with unstructed data. Where credit is due
Every enterprise has overdue challenges in maintaining good data hygiene, ensuring faster retrieval of data when it's needed. This put Asian Development Bank in the situation where there's more than half a century of information scattered across various locations, particularly Sharepoint and individual computer devices.
"We have a specialised role, called the front desk where we ask for domain experts and where can I get this information?" — Anon. Project Consultant
To get the information the consultants needed, it requires good knowledge and network to know who to ask for a chat to get the necessary information on the project they are working on.
ADB Navigator as we know today
Wearing The Users Shoes
I'll keep this short since I wasn't fully involved while interacting with the consultants during the initial interviews. Though I briefly participated the design workshop as one of the facilitator, co-designing the ideal experience they wanted while interacting with the Generative AI.
3 key learnings while ideating the ideal experiences:
Use familiar interfaces from ChatGPT, Perplexity for fast adaptation of the new tool
With conversational AI interaction, it will quickly become a garbage-in, garbage-out situation. How can we store prompts that generates the golden response? Golden response means a perfect score response generated by the AI.
Other than searching and probing the knowledge, can we generate project documents with the information I asked for.
ADB Navigator
Perplexity was popular with the consultants, due to the ability to cite sources used for generating the infromation, and ablility to probe further with never thought of prompts. It became the benchmark while designing ADB Navigator.
"No, ChatGPT interface isn't enough. It is too plain and I want to think less when I start investigating a domain topic." — Subject Matter Expert Director
Release 1 of ADB Navigator
To ensure we are capturing the ideal product experience, a Figma prototype was quickly built with painstakingly structured the generated output as it would be in the real world, imitating the AI interaction and probing of the infromation.
After 3 crucial rounds of iterations, ADB Navigator have some unique value that could make the consultants' information seeking more engaging:
Key personnels contact are made available, based on authoring level of the information, for a more classical communication.
Related knowledge presented to providing more reading materials that have topics or domain information that are related to what the consultants are searching for.
Impact, Measured With Satisfaction
"Whoever is the user experience designer behind this product, please help me convey my gratitude for having such an intuitive experience while using the product." - Sr. Consultant
ADB Navigator has one product metric to be measured, how fast a consultant can get the information he/she may need?
Prompt library helped AI illiterate consultants to get quality responses by using shared golden prompts, reducing poor quality responses by 40%.
With less than 5 clicks, credible related information can be used as part of the project reports.
To date, more than 3,000 internal and external consultants working at ADB are using ADB Navigator as their source of truth.
ADB Navigator for mobile, Google Playstore
Today, the product has now scaled to have a mobile app, focusing on a voice-to-text input model, bringing refined AI experience to the next level for enterprise use case.
Conclusion
The ADB Navigator journey demonstrates how thoughtful AI design can transform information chaos into structured knowledge at enterprise scale.
By focusing on familiar patterns, quality prompts, and contextual information, we delivered a solution that not only solves the immediate problem of information retrieval but also enhances the collaborative culture at ADB. This project exemplifies my approach to designing AI experiences: simplifying complexity, prioritizing user context, and measuring success through tangible outcomes.
Notes
Mock data replaces actual information to safeguard confidentiality. The value delivered to more than 5,000 project consultants is 100% authentic
ADB Navigator goal is to be the source of truth for the massive information generated for more than half a century.
As an independent contributor AI Experience Designer in the product team, I've designed the digital product with heavy conversational interaction, allowing consultants to get factual information for their projects.