KERALA RISING · FROM EXPORTING PEOPLE TO EXPORTING IDEAS What if the revenue never materialises? Then the institution needs state operating support for longer before self-sufficiency, the downside scenario, but it still works, because the cost side is modest and the AI is a rounding error throughout. The honest risk is on the revenue ramp, which is where a reviewer should push. See Chapter 16. On ethics and trust Is this surveillance? No. It is opt-in, recognitiononly, never punitive, blind to clinical and affiliation data, and governed by the DPDP Act, with a public corpus rather than a hidden database. The architecture is the opposite of a surveillance system. See Chapter 22. What about the digital divide? Leaderboards rank on participation rate, not volume; field operators assist logging for households without smartphones, and the on-ramp runs through the institutions that already reach the poor. Equity is a measured, audited parameter. See Chapters 18 and 22. What if the AI is wrong? It will be sometimes; the design assumes so. Verification is conservative, human-audited on a sample, inspectable, and publicly correctable. The mitigation is auditability, not perfection. See Chapter 4 and the chapter on what the AI can and cannot do. 340