AI & Ethics
Explainability, model governance, human-in-the-loop review, bias monitoring, and decision traceability — the accountability layer that makes AI safe for institutional decisions. PRAGATI is human-guided AI: the system surfaces evidence, your leadership acts on it.
Explainability by design
Every score traces to a transparent formula; every answer shows its sources, confidence level, and known limitations. No black-box outputs.
Human-in-the-loop
High-stakes recommendations are flagged for human review and never auto-actioned. PRAGATI recommends — your leadership decides.
Bias & coverage monitoring
Average confidence by region surfaces where the model is least certain, so those markets receive extra scrutiny rather than false precision.
Decision traceability
An immutable decision log records every AI-assisted output — model version, confidence, sources, and human-review status — for audit and compliance.
Model governance
A registry documents each model's purpose, inputs, evidential basis, and human-review policy. No undocumented AI in production.
Grounded, not generative
DRISHTI answers only from your institution's connected data. It does not fabricate or infer beyond what the evidence supports — and says so when confidence is low.
Built for institutional accountability
University leadership, boards, and accreditation bodies expect transparency from any system influencing institutional decisions. PRAGATI is designed to pass that scrutiny — every output is traceable, every model is documented, and every high-stakes recommendation is held for human review before action.
All AI outputs in PRAGATI are derived from your institution's own connected data (Banner, Slate, Terra Dotta, ViaTRM, Salesforce). PRAGATI does not train on institutional data and does not share data between institutions.