Why collateral verification decays after sanction and how AI re-reads EC, litigation and mutation records quarterly across the whole book, grading deterioration for credit teams.
A bank's collateral verification is a snapshot: on the report date, title was marketable and the EC was clean. The borrower then holds the asset for years. In that time a second charge can be created, a revenue attachment recorded, a partition suit filed, or an apparently released mortgage revived through fraud. Standard practice catches this at renewal, if at all, and most often at default, when recovery options have already narrowed.
The gap is structural: re-verification by hand costs nearly as much as the original diligence, so it is rationed to stressed accounts. The healthy book runs blind.
For each secured parcel, the monitor periodically re-pulls the encumbrance record and scans for new entries after the charge date, checks cause lists and case-status sources for the borrower and the property's survey numbers, and watches mutation and patta changes that suggest a transfer or succession event. New facts are diffed against the baseline report and graded: a fresh second charge to a regulated lender reads differently from a revenue attachment.
The economics only work because the reading is mechanical. A human team re-checking a 5,000-account book quarterly is a department; a machine doing the same is a schedule.
Graded alerts map to the credit playbook: Low findings log to the file; Medium triggers a borrower query or covenant letter; High, an unauthorised charge or an attachment, escalates to legal for protective filing while the lender's position is still senior and the borrower still cooperative. The asset that quietly accumulated three adverse entries over two years stops being a surprise at NPA classification.
Portfolio-level, the same data ranks the book by collateral deterioration, sharpening provisioning and inspection priorities with evidence rather than vintage heuristics.
Tamil Nadu's records suit monitoring: ECs are obtainable per survey number and period, TNRERA filings are public for project exposure, and court data is increasingly searchable. We run collateral verification at origination, the full 30-point read with legal validation, and continue the same checks as a quarterly watch through the life of the loan.
For LAP and SME books, where collateral quality varies most, the watch typically pays for itself with the first early-caught charge.
Quarterly re-pulls of the encumbrance record and litigation status are practical once reading is automated. At minimum, re-verify at every renewal and before any enhancement, on the whole book rather than a sample.
New charges and second mortgages, revenue or court attachments, lis pendens entries, partition or succession events, and unauthorised transfers. Each appears in public records well before it appears in a default file.
No. It extends it. The origination report is the baseline; monitoring diffs new entries against that baseline and grades them, so the original verification keeps earning its cost through the holding period.