In short
The same customer complaint can reach an Indian bank through four separate channels: the NCRP and I4C chain, a law enforcement email, the customer's own email, and the grievance route. Each arrives in a different format with no shared identifier.
So the first task of the day is building a queue rather than working one. In one deployment an analyst then touched five systems per case, and response time ran at about three days before any of it was decision-making.
Four channels, one customer, no shared reference
A defrauded customer rarely uses one route. They report the fraud, they call the bank, they email, and if nothing moves they escalate through the grievance route.
Meanwhile the complaint they filed has generated a case that reaches the bank through the NCRP and I4C chain, and a law enforcement agency may separately send an information request about the same transactions.
That is one event producing four pieces of work in four formats, landing with different teams, with nothing tying them together automatically.
Nobody designed this. It is what happens when each channel was built to solve its own problem.
Even the complaint itself has more than one front door. The Ministry of Home Affairs SOP of 2 January 2026 recognises registration through the NCRP portal, the 1930 helpline, police stations, and banks registering complaints on NCRP.
What each channel actually asks of the bank
| Channel | What arrives | What it needs back |
|---|---|---|
| NCRP and I4C chain | A structured complaint with a reference | A response in the prescribed format, and a hold where the money is still there |
| LEA email | A request for information, action, or a notice | A fixed-format reply with statements and KYC attached |
| Customer email | An unstructured description and a lot of urgency | A clear answer a person can understand |
| Grievance route | An escalation that already failed once | A resolution and a record of why it failed the first time |
Read the right-hand column. Four different outputs, four different standards of proof, and the same underlying facts behind all of them.
The law enforcement channel needs its own section
The law enforcement channel is consistently underestimated, so it is worth the specifics.
In one Indian banking operation, seven distinct scenarios were about law enforcement instructions, accounting for 2,413 cases out of 16,736, or 14.4% of the queue.
The instruction itself turns up in three different places: matched through the bank's API in 1,660 cases, found on the portal in 302, and buried in older email threads in 307. An analyst checks all three because there is no way to know in advance.
A further 120 cases had no matching instruction at all, and 11 carried a freeze code with nothing behind it. Those are the cases that take longest and carry the most risk.
Patched turns messy, manual, spreadsheet-driven operational processes into reliable AI-assisted workflows with human review where needed.
Why the freeze creates a fifth queue
Acting on a law enforcement instruction is not the end of the case. It creates the customer conversation that follows.
Frozen-account cases were 3,755 of 16,736 in the same operation, or 22.4% of everything, spread across 37 distinct scenarios.
That queue is generated by someone else's decision and staffed entirely by the bank, and on most org charts it has no owner at all.
What a single intake standard changes
The fix is not a new portal. It is one working queue assembled across the channels you already have, with the evidence attached before an analyst opens anything.
That means the same customer's four arrivals get recognised as related, the transaction history and KYC records get pulled once rather than four times, and the case reaches a person as a prepared recommendation rather than a starting point.
The measurable effect in one deployment: response time moved from about three days to under an hour, analyst hours per case fell roughly 75%, and turnaround time came down 40%.
Where the analyst's time moves to
Figures from one Indian banking deployment, 2026.
Between 50 and 60% of cases are handled end to end. The rest still involve a person, but the work per transaction drops from about two minutes to under one, because the evidence is already there.
How to see this in your own operation
Take one customer who escalated last quarter and trace every piece of work their case generated across all four channels.
Count the teams involved, the systems opened and the number of times the same transaction history was retrieved. Most banks running this exercise find the same evidence pulled three or four times by three or four people.
That duplication is the cost of four front doors and no shared queue - and it is the cheapest thing on this list to fix.
Why deduplication is harder than it sounds
The obvious fix is to match the four arrivals to one customer and work them together. Doing that reliably is harder than the idea suggests.
There is often no shared reference. The complaint carries its own identifier, the law enforcement request refers to a transaction, the customer email refers to neither, and the grievance refers to an internal ticket.
Matching therefore happens on the underlying facts: the account, the transaction, the counterparty, the date range. That is a retrieval problem before it is a matching problem, and retrieval is where the case time already goes.
Which is why this rarely gets solved by a rule in a ticketing system. The information needed to match is sitting in the five systems nobody has opened yet.
What good looks like from the customer's side
A single intake standard is worth checking against the person on the other end, because that is the part the org chart hides.
Today a defrauded customer can be told four different things by four parts of the same bank within a week, none of them wrong and none of them complete. The frozen-account conversation is the sharpest version, and those cases were 22.4% of one queue.
A single working queue is not only an efficiency argument. It is the difference between a bank that sounds like one organisation and one that sounds like four.
Sources
- Scenario breakdown from one Indian banking operation, 2026: 16,736 cases, 2,413 LEA cases, 3,755 frozen-account cases.
- Patched data points, 2026: five systems per case, 3 days to under 1 hour, roughly 75% analyst hours.
- Ministry of Home Affairs, Standard Operating Procedure for NCRP-CFCFRMS, Custody, Restoration of Money and Grievance Redressal, 2 January 2026.
