Banking & Capital Markets Systems · Chapter 9 of 13
Risk, AML, and why regulatory reporting is a lineage problem
OFSAA, the detection layer, and the RBI mandate that is really about data extraction.
These two sit adjacent on the stack and get merged in job titles, and separating them is the most useful thing this chapter does.
Detection — the AML layer
Oracle OFSAA spans asset-liability management, liquidity risk, funds-transfer pricing, credit risk and expected credit loss, stress testing and regulatory reporting. Oracle Financial Services ranked #3 in the 2025 Chartis RiskTech100 and #4 in 2026 — a real, dated credential, though a global one.
The “very large India footprint” claim could not be quantified. Hundreds of open OFSAA roles concentrated in Bengaluru are suggestive of demand, but job-board density is not evidence of production deployments at named banks. Treat it as directional. And for NICE Actimize, Oracle FCCM and Wolters Kluwer OneSumX, no named Indian bank could be confirmed at all — the categories are certainly present, the deployments are not public.
Lineage — the reporting layer
RBI's Automated Data Flow, introduced around 2010, required banks to extract returns automatically from their core systems, maintain central metadata repositories, and submit in XML or XBRL with no manual intervention, across up to 250 distinct returns. Its successor, CIMS, took over quarterly statement submission from the quarter ending March 2024, added digital lending app reporting in May 2025, and internet and mobile banking returns from September 2025.
The distinction to make in an interview
AML is a detection problem. Regulatory reporting is a lineage problem. They are not variations on the same thing.
Transaction monitoring asks is this behaviour anomalous? — a modelling question, with false positives as the central operational cost, and it lives with the alert-fatigue problems every detection discipline has.
The ADF-to-CIMS mandate asks something completely different: can you produce the same number the same way every time, straight through from the ledger, with a trail that survives an auditor? By construction that is about extraction, standardised taxonomy and eliminating manual touch — the hard part is lineage, not detection.
Candidates who have only worked on one of these frequently assume the other is easier. Saying which one you have actually done, and what was hard about it, is a stronger answer than claiming both.
Note that CIMS is still being rolled out category by category, which means ADF may be fully retired by the time you read this. It is a live migration, not a settled state.
Check yourself