EASE 9.0 · Risk & Resilience · Technology Modernization & RenewalRBI Stopped India’s Biggest Card Issuer Cold for Eight Months
Legacy architecture was built for overnight batches.
That world is over, and the infrastructure has to change with it, without betting the core on a single cutover.
Most PSU banks run on technology built ten to twenty years ago: reliable, proven, and designed around the overnight batch. Transactions queue during the day and are processed, checked, and settled at night. That architecture made sense when customers expected next-day answers. They no longer do.
EASE 9.0 assumes something different: microsecond latency, compliance gates built into the decision itself, active-active data centres, and core AI stacks running on private cloud. Technology Modernization is not one category among sixteen. It is the prerequisite for progress on almost every other one.
The Path to Progress
Banks that get this right do not replace the core in one migration event. They run the new platform and the legacy core in parallel, in five phases, over roughly two years.
Replacing the core in one migration event invites an outage exactly when the bank can least afford one.
The Contrast Between a Phased Migration and a Big-Bang One
This is a documented failure mode, not a theoretical one, and it has already happened to India's largest private bank. Between 2018 and 2020, HDFC Bank suffered a string of digital and net-banking outages as customer volume outgrew infrastructure that had not been modernized to carry it. In December 2020, RBI barred the bank from launching any new digital initiatives and, more consequentially, from issuing any new credit cards at all, until the underlying technology was fixed. The credit-card freeze held for eight months; HDFC Bank's card base shrank from 15.38 million to 14.82 million over that period even as competitors kept issuing, and the wider Digital 2.0 restrictions were not fully lifted until March 2022, fifteen months after the embargo began. The ambition of HDFC Bank's digital strategy was never the problem; the infrastructure underneath it was.
Source: RBI directive, December 2020; RBI lifts restrictions on credit-card issuance (August 2021) and Digital 2.0 (March 2022), as widely reported.
The pattern is not unique to India. In April 2018, TSB Bank in the UK moved five million customer accounts to a new core platform in a single weekend cutover; customers were locked out for weeks, fraud losses spiked into the gap, and the FCA and PRA jointly fined TSB £48.65 million for inadequate migration planning and governance. Commonwealth Bank of Australia took the opposite approach, replacing its legacy core product-by-product over roughly five years, then running an 18-month phased migration of its core banking system to cloud in 2024–2025 with no comparable disruption. Different regulators, different banks, the same underlying lesson: infrastructure debt eventually becomes a regulatory event, not only a customer-experience one.
Sources: FCA/PRA enforcement findings on TSB's 2018 migration, as widely reported; Commonwealth Bank of Australia newsroom, core banking cloud migration case study (2026).
The Business Case
Where AI Agents Fit
No customer ever sees a modernized core directly. This category is pure infrastructure, but it is the infrastructure every named agent in this guide runs on. Every agent architecture in this guide starts with a Sense layer, the component that assembles a live, unified view of the customer from core-banking events, app behaviour, and branch signals as they happen. An agent cannot decide in real time on data that only refreshes overnight, so this is the one category where being an enabler means being literally load-bearing: a Sense layer built on a batch feed can only describe the customer as of last night's close, and every downstream agent inherits that stale picture.
What Banks Get Wrong
- The big-bang cutover. Replacing the core in one migration event, rather than running old and new in parallel, invites an outage exactly when the bank can least afford one.
- Treating it as an IT project. Without a named business owner accountable for the categories a modernization programme unlocks, funding gets renegotiated the first time a competing priority appears.
- Modernizing the channel before the decision. A faster app wrapped around the same batch-driven engine ships a better interface on top of the same slow answer.
The Skeptic's Answer
A phased, two-year migration invites a fair worry: that it becomes a way to never finish, a multi-year IT programme that survives three budget cycles, two CTOs, and a change in vendor, while the fintech competitor ships a new feature every quarter in the meantime. TSB's cutover was fast and it still failed; the fear that slow is just a different way to fail is not unreasonable.
Fair, on the record alone. What that argument misses is what "phased" actually buys, not only what it avoids: CBA's teams shipped real, customer-visible improvements at the end of Phase 1 and Phase 2, a modernized data layer and a real-time decisioning engine, years before Legacy Decomposition was even scheduled. A phased migration is not slow value; it is value delivered in an order that never requires betting the whole bank on one weekend. The two-year timeline is the cost of never having to explain a TSB-style outage to a regulator.
Modernize Alongside the Core
Appice's platform runs this exact five-phase journey without forcing a bank to gamble its core banking system on a single cutover. The decisioning layer deploys on the bank's own infrastructure and connects to the legacy core through the integration patterns most PSU banks already run today, so Phase 1 and Phase 2 can start almost immediately, proving value on one function before Channel Migration ever becomes a board-level decision.