> ## Documentation Index > Fetch the complete documentation index at: https://docs.superadmission.com/llms.txt > Use this file to discover all available pages before exploring further. # The built-in human reviews > Where an actual person is required before something becomes final, and the principle behind why those specific points and not others. The overview page for this section already said this plainly: PraveshAI doesn't publish an allotment on its own, doesn't make a final call on an uncertain document, and doesn't resolve a grievance by itself. This page is about where those checkpoints actually sit, and why they sit there specifically. ## The Principle The rule used to decide where a human checkpoint belongs is simple: anywhere a decision could materially change whether a specific student gets a specific seat, a person reviews it before it becomes final. Anywhere a decision is just removing repeated manual work, checking a document's format, syncing a deadline across two dashboards, no outcome-changing judgment is being made, so no checkpoint is needed there. ## Where the Checkpoints Actually Are Only documents scoring highly on the automated check clear without a person looking at them. Anything less certain goes to a reviewer. The exact scoring behind this is covered on How a Document Actually Gets Verified. A completed match never reaches a student automatically. It goes to the authority for review and requires their explicit sign-off first, covered in full on How a Counselling Authority Actually Runs a Round. Every dispute is resolved by a person, categorised and routed to whoever is positioned to actually decide it, never closed automatically. The full path is covered on How a Dispute Actually Gets Resolved. The most disruptive actions available require explicit authorisation, and a rollback specifically requires more than one person to approve it, covered on The Safety Levers Behind Every Round. ## Why This Matters More Than It Might Seem A seat allocation decision affects a real student's future, sometimes irreversibly, once they've acted on it. That's a fundamentally different kind of decision than syncing a deadline reminder, and it's treated differently on purpose. This is also the specific reason PraveshAI is built to stay inspectable rather than opaque, in line with the Safe and Trusted AI focus of the IndiaAI Mission, mentioned on Superadmission as Digital Public Infrastructure. A system making decisions this consequential has to be one where a person can always ask why, and get a real answer, not just a result. That "real answer" part, how every decision, human or automated, gets recorded so it can actually be explained later, is covered next.