AI LLM Chat Assistant for Data Analytics Products
World Bank staff needed a faster, more accessible way to extract insights and visual reports from complex datasets without relying heavily on analysts or navigating multiple tools.
Challenge
The World Bank provides a suite of data analytics products such as reporting tools and data explorer platforms, to help staff and stakeholders access and analyze global financial, social, and economic datasets.
Provide quick, natural-language access to datasets.
Generate visualizations (charts, graphs, dashboards) alongside text-based outputs.
Deliver domain-specific insights and summary reports.
Support sharing of insights with higher management in a clear, professional format.
Complex navigation: Finding the right dataset across multiple tools required deep knowledge of data structures.
Time-intensive reporting: Generating management-ready insights meant downloading, filtering, and synthesizing raw data manually.
Limited accessibility of insights: Non-technical users struggled to derive value from the data without analytical expertise.
Prototype
Solution
We designed and prototyped a Large Language Model (LLM)-based chat assistant embedded into the World Bank’s reporting and analytics products.
Users ask questions in natural language (e.g., “Show me GDP growth in Sub-Saharan Africa for the last 5 years”)
Conversational data querying
Dataset discovery
The assistant suggests relevant datasets available in World Bank repositories.
Instant insights
The assistant suggests relevant datasets available in World Bank repositories.
Pre-trained prompts for common World Bank use cases (e.g., poverty rates, climate finance, project performance).
Conversational data querying
This solution reduced friction, democratized access to complex datasets, and accelerated insight generation for decision-making.
My Role
As Product Designer, I led the end-to-end design process—from discovery to delivery:
Design Strategy & Concepts:
Defined vision, assistant capabilities, and integration roadmap.
Research:
Conducted user interviews with analysts, economists, and policy experts.
Workshopping Use Cases:
Facilitated design thinking workshops to prioritize scenarios.
Wireframes & Prototypes:
Built conversational UI flows, dashboard mockups, and high-fidelity prototypes.
Visual Design & Specs:
Delivered production-ready specs and visual comps.
Usability Testing:
Validated workflows with target users, iterating based on feedback.
Primary Research
I created a journey map and visualized the complete integration of AI capabilities into the World Bank's operational reporting workflow, demonstrating how AI can enhance decision-making through conversational interfaces and automated insights generation.
The AI LLM Assistant integration into the World Bank Operations Reporting Portal, showing the complete user journey from login to automation setup.
User profile and empathy maps
Working with a Internal Research team, we came up with a various users from technical power users to non-technical executives, each with distinct needs for AI assistance. we tried maintaining consistent value delivery across different organizational roles and responsibilities.
Impact
Accelerated insight generation
Reduced time to find and analyze data from hours to minutes.
Non-technical users were able to generate management-level insights without relying on analysts.
Increased adoption
The assistant’s ability to cite sources and generate visuals enhanced confidence in outputs.
Improved trust and collaboration
The design blueprint set the stage for scaling the assistant across other World Bank tools.
Cross-product value
Metrics
Time Saving
60% faster from query to usable insight.
User satisfaction
85% of pilot users reported higher confidence in accessing data.
70% of early testers used the assistant weekly for reporting tasks.
Adoption
Analysts reported a 30% reduction in ad-hoc data requests from management teams.
Efficiancy