AI Methodology
How we build insights, the principles we hold ourselves to, and the limits we are honest about. We treat AI as a tool that calculates answers from real data, never one that invents them.
Last reviewed: May 26, 2026
Our principles
Six commitments shape every design decision behind Amrut AI.
Grounded in real data
The model can only state numbers and facts a verified data source has returned. No free-form invention.
Cited inline
Every claim is paired with the underlying source row so you can audit where it came from.
Honest about gaps
When a signal is unavailable, Amrut AI says so plainly instead of filling the gap with guesses.
Refreshed continuously
Different signals refresh on different cadences. Insights regenerate when the underlying data shifts.
Privacy-respecting
Only public, ethically obtained signals are used. No personal data, no logged-in surfaces, no scraping behind authentication.
Human-reviewed
Our data team reviews the methodology, prompt templates, and sampled edge cases every release cycle.