Where I work between the counter and the board

I treat a rural bank merger as an operating problem before it is a cultural one. The eleven institutions that formed Xuantang ran on the same provincial core system but had different authorisation limits, voucher forms, opening hours, cash routines and ideas of what a supervisor could approve. A teller transferred from one county to the next could not do her job on her first morning. I map those differences process by process, decide with the county branch presidents which practice survives and publish each decision, including the ones where a smaller county's way was better. Local exceptions the merger has not yet removed go on a register with an owner and an end date, and I report that register to the board's risk management committee. I also watch the measures that show strain early: counter errors in the months after a change, complaints by county and transfer requests from township staff.

I have centralised bank back offices twice, at a national joint-stock bank between 2008 and 2014 and at Xuantang since 2023. The technical work is familiar: scanned vouchers, remote authorisation, central account-opening review and standard cut-off times. The judgement lies elsewhere. I decide how many sites each process needs to survive an outage, which processes must stay close to the counter and how the counter will see the benefit within weeks rather than years. I measure a centralisation by the counter's waiting time and the error rate on returned vouchers, both published weekly by outlet. At Xuantang I chose three processing sites over one, which cost about RMB 26 million a year of planned savings. In July 2025, when flooding closed our southern site for three days, its work moved to Dongpu within two hours. I report an annual failover test to the risk management committee.

Since the anti-telecom and online fraud law took effect in December 2022, banks in China have had clear duties to identify risky accounts, limit how they can be used and cooperate with the public security organs. I run the operational side of those duties at Xuantang: transaction monitoring alerts, counter interventions when an older customer is being coached to move money, and the handling of restrictions. Policy and rule approval sit with our head of legal, compliance and anti-fraud; restrictions instructed by the police follow their own process, which I do not second-guess. My particular concern is restrictions created by our own rules. Each one is a judgement about a customer, and I insist that it has an owner, a review clock and a release route the customer can actually use. I report release times to the risk management committee beside fraud losses, because either figure alone flatters the bank.

In a prefecture where 1.4 million of our customers are over sixty, the outlet network is a service commitment before it is a cost line. I plan closures and openings with three tests: whether another outlet lies within two kilometres, whether the receiving outlet can handle its busiest pension day, and whether nine in ten customers over sixty-five can reach it within forty minutes by bus. I pair each closure with something that stays: a mobile service vehicle, extended village service points or staff who move with their customers for six months. Inside the outlets I run what the regulator calls age-friendly service: priority counters, large-print forms, unhurried help with mobile banking and a standing instruction that no older customer is turned away for want of a smartphone. Counter staff are also our best defence against scams aimed at older people, and they are trained and measured for it.

I do not make credit decisions. Credit authority at Xuantang sits with county loan officers and an approval chain that runs to our chief risk officer. What I own is everything around the decision that makes a loan slow or error-prone: document collection and checking, contract preparation, collateral registration, disbursement and the post-loan file review. In 2024 we moved those tasks into a central credit operations unit of 140 people, largely staffed from the merged back offices. Median turnaround for small-business and farm loans fell from twenty-one working days to eight, and documentation defects found in post-loan review fell from 31% of files to 9%. I am careful to keep the county loan officer as the borrower's contact and the person who makes the credit judgement. Village-wide credit assessment depends on local knowledge that no central team can collect from a file.