Make judgment cheap enough to travel.
The commercial case is not automation for its own sake. It is rigorous evidence on tickets that arithmetic currently excludes.
The floor is arithmetic, not intent.
Drag the fully-loaded cost of one rigorous appraisal. The curve shows the smallest ticket that can carry that work under this illustrative bank hurdle.
A large ticket can absorb rigorous judgment today. That is not the rural growth strategy.
Lower the cost floor and the same evidence discipline can reach a new-to-credit borrower.
The cluster’s simulated average ticket is ₹1.63 lakh. Lowering the cost floor brings that evidenceable population into economic range.
A task table, not a headcount promise
Human authority stays on conflicts, exceptions and the decision. The agent workforce expands the evidence work around that decision. “Rarely done today” is marked rather than quietly counted as productivity.
| Credit task | Human alone | Human + workforce | What changes |
|---|---|---|---|
| Research schemes and regulation | 2.0 h | 0.3 h | repeatable |
| Reconcile conflicting sources | 1.5 h | 0.5 h | judgment retained |
| Financial spreading | 2.5 h | 0.4 h | repeatable |
| Site-visit note synthesis | 1.0 h | 0.2 h | rarely done today |
| Scheme eligibility mapping | 1.0 h | 0.2 h | repeatable |
| Draft CAM | 4.0 h | 0.8 h | maker-checker |
| CP tracking | 0.8 h | 0.1 h | repeatable |
| Post-disbursement EWS scan | rarely done today | 0.2 h | new coverage |
| Renewal review | 3.0 h | 0.6 h | judgment retained |
| Cluster origination | rarely done today | 0.5 h | new coverage |
Simulated workshop observation. Not a public benchmark and not a claim about every bank.
The 90-day pilot measures whether the architecture earns this claim. It is not measured today.
Both baseline values are placeholders for the bank’s own measured baseline, not sourced industry facts.
Measure the cost floor, officer throughput and quality blind-scored by the bank’s own credit committee. Keep the numerator and denominator visible.