Decisions that cost the bank something, and why I made them

Re-underwriting pandemic restructurings before the relief ended

When I took office in June 2021, Rp 11.8 trillion of our loans, about a fifth of the book, had been restructured under OJK's pandemic relief. The relief allowed a restructured loan to remain classified as current, which was right in 2020. By late 2021 it had become a comfortable fog. Our MSME business expected most borrowers to recover; our collectors, asked directly, thought perhaps a third would not. The relief would end, and whatever we had not recognised by then would arrive at once.

In November 2021 I proposed three things to the Board of Commissioners: re-underwrite every restructured borrower above Rp 500 million, about 4,900 of them, with a site visit; raise expected credit loss coverage on the restructured book from 6% to 14% during 2022; and recommend a payout of 15% of 2021 profit instead of the 45% the bank had paid for a decade. Our director of MSME banking, Hadi Wibisono, argued that so many visits would alarm borrowers and that rolling the loans forward was kinder. The family holding company serviced its own borrowings from our dividend, and Benedict Tjandrakusuma, then still vice president director, said so plainly.

The commissioners approved after two meetings, independent members carrying the argument. We gave up a year of earnings and most of a dividend: 2022 net profit fell to Rp 710 billion from Rp 1.14 trillion, and our share price fell 17% in the following quarter.

Visiting in pairs, our officers, analysts and collectors sorted borrowers into 3,350 who had recovered, 980 needing a second restructuring backed by fresh owner equity or collateral, and 570 whose loans we sold or wrote off.

When the relief ended in 2024, our pandemic-restructured book was Rp 2.9 trillion. Gross non-performing loans peaked at 4.6% in mid-2023 and stood at 2.8% in June 2026; loan at risk fell from 21.4% at the end of 2021 to 10.2%. Credit cost fell to 1.3% in 2025.

Our dividend policy now links the payout to thresholds published in advance. What I would do differently concerns fourteen Bali hotel loans worth Rp 1.3 trillion. I gave them time because tourism was returning; we sold two a year later at deeper discounts. I would also rewrite our first letters to borrowers, which read like legal notices and drew over 300 complaints.

Closing urban offices without leaving rural markets

In 2023 our 364 offices accounted for 41% of operating costs. In Jakarta, Surabaya and Semarang, 78% of retail transactions already ran through our app. In South Sulawesi, Central Kalimantan and most of our Sumatran districts, traders still deposited cash daily and borrowers expected a loan officer within reach. A consulting review recommended closing 120 offices, ranked by cost per transaction, for an annual saving of Rp 410 billion.

I proposed a narrower plan. We would close or merge 58 urban offices that sat within 1.5 kilometres of another Kartadana office or where more than 80% of transactions were already digital. We would convert 41 rural cash offices into lending posts staffed by loan officers, move cash services to 2,300 additional Laku Pandai agents and close no rural office. Our finance director, Melati Hasibuan, argued for the full programme, pointing out that the difference in savings exceeded our annual technology budget. Our director of digital banking, Arief Kusnadi, saw agents as a halfway house that would slow customers' move to the app. An independent commissioner asked how many deposits we would lose.

We accepted a saving of Rp 172 billion a year instead of Rp 410 billion and spent Rp 64 billion on agent devices, lending-post refits and training. Each closure was assessed against deposits, loans, complaints and the nearest alternative point of service, and customers received three months' notice.

By the end of 2025, 91% of deposits held at the closed offices had stayed with us. MSME loan origination through the converted lending posts rose 23%, because loan officers now spent their days with borrowers instead of behind counters. Agents handled 2.2 million transactions a month, and monthly active app users rose from 0.9 million to 1.6 million. The network now has 306 offices and about 6,900 agents.

Every office now has an annual review against the same four measures, and no closure is approved without a named service alternative. What I would do differently is plain. About 4,100 retired customers drew their pensions through accounts at the urban offices we closed. Many could not use the app, had to travel further and for two months queued at the receiving offices. We should have arranged pension payouts through agents, and home visits for the frailest, before the closures rather than after the complaints.

Adding friction to stop scams before customers paid

During 2023, 6,400 of our customers reported losing a combined Rp 87 billion to scams. Most followed a pattern familiar across Indonesia: a chat message carrying a file disguised as a parcel receipt or a wedding invitation, which installed software that read the customer's screen, or a caller posing as bank staff who talked the customer through a transfer. Our fraud systems saw a known customer, on a registered phone, making a payment they had authorised. In the same year 11,200 accounts at Kartadana received scam proceeds, many opened by people paid to lend their identity.

In February 2024 I approved four changes. The first transfer above Rp 5 million to a new beneficiary, made within twenty-four hours of registering a new device, would be held for four hours while we called the customer. The app would stop working if an unknown application held screen-reading or remote-control permissions. Account opening would add biometric matching against population-registry data and device checks. And we would publish a framework for reimbursing customers where our own controls had failed.

Arief Kusnadi, our director of digital banking, opposed the hold. He expected ratings to fall and customers to move payments to competitors, and for a while he was right on both counts. Our MSME business warned that merchants paying suppliers would object. Our legal team feared the reimbursement framework would read as an admission of liability.

The costs were visible. Our app rating fell from 4.5 to 3.9 over five months, transfer volumes dropped 6% for two months and 3,100 merchants complained in the first month. We reimbursed Rp 14.2 billion across 2024 and 2025.

In 2025 customers reported Rp 26 billion of scam losses across about 2,500 cases. Seventy-one per cent of the transfers we held were cancelled by the customer after our call. Accounts at Kartadana receiving scam proceeds fell to 3,900, and the app rating recovered to 4.4.

Scam losses and reimbursements now go to the Risk Monitoring Committee every quarter, beside the complaints the controls generate. What I would change is the first design. We applied the hold to long-standing business accounts as bluntly as to new retail customers, and we should have tiered it from the start. I would also have explained the change before it began, not apologised afterwards.

Declining a related-party structure proposed by our own shareholder

In February 2025 our controlling family's property business, Wisesa Properti, asked us for a Rp 1.1 trillion construction facility for a township near Karawang in West Java. Lending to related parties is capped by Indonesia's legal lending limit, which for us stood at about Rp 1.3 trillion, and our existing related-party exposure was Rp 640 billion. The request proposed lending instead to a joint venture in which Wisesa held 49% beside an unrelated developer, arguing that the venture was not a related party. Our director of corporate banking supported it for its Rp 38 billion of first-year fees, and the project was sound on its own terms.

I did not accept the structure. The family's businesses would contribute the land, appoint the project director and stand behind cost overruns, and I concluded that the venture's credit was, in substance, the family's. Treating it as unrelated would have put us at 135% of our limit in all but name, and I did not believe that would survive an examination. I proposed a facility of Rp 420 billion, within the limit, on terms checked against comparable unrelated developers and decided by independent commissioners with family commissioners recused.

Benedict Tjandrakusuma, our President Commissioner, argued that the joint venture was genuinely independent and that the bank was treating its founding family less favourably than a stranger. One independent commissioner went the other way and wanted to decline outright. Both positions were minuted.

We took an external legal opinion and put the proposal through the Audit and Risk Monitoring Committees. Benedict recused himself, and our independent commissioners approved Rp 420 billion. The developer raised the balance from two other banks. We gave up roughly Rp 95 billion of fees and interest over three years, and for some months my relationship with our largest shareholder was formal rather than easy.

Related-party exposure now stands at 8.2% of capital. Any related-party transaction above Rp 25 billion is decided by independent commissioners with family commissioners recused, and we report the total to the general meeting each year.

What I would do differently is the sequence. That rule should have existed in 2021, before any request arrived; writing it in the middle of one made it look personal. And Benedict first learnt of my recommendation from the credit memorandum. I should have told him myself on the day I reached it.