An evidence-backed website mapping how 12 Bengaluru public services depend on records and procedures controlled by other departments.
I built the atlas for Bengaluru residents trying to understand connected tasks such as property-tax changes, water-account transfers, and electricity connections. I asked the AI to research each service through official sources, public workflow observations, and redacted citizen accounts, then organize the findings into structured evidence ledgers. The AI produced the data-driven website and automated checks; I set the scope, evidence rules, and requirement to show uncertainty instead of guessing. The finished atlas covers 12 services and identifies 47 places where public documentation stops before the citizen’s task does.
We added a separate “documentation debt register,” but it repeated information already present in each service entry and made the site feel more abstract. We removed it and used a filing-cabinet view to show existing documents and missing procedures more directly.
I began with a bounded product and research method, then split the work into service-level research, independent audits, and interface passes. Later prompts focused on one problem at a time, including citation quality, plain-language summaries, navigation, and removing repeated explanations.