Swiss Banks Had a Record Year. So Why Are They Less Efficient?
21
Overview
Swiss private banks grew assets last year and still got less efficient. Most run five transformation programs at once: operating model, data, AI, technology and skills. The order matters more than the budget.
Year
2026
Industry
Financial Services
Challenge
Most banks run five transformation programs at once. The order matters more than the budget. Swiss private banks had a record year for assets and a worse year for efficiency. KPMG and the University of St. Gallen studied 68 banks and published the results on 24 June 2026. The median cost-income ratio rose from 75.6% to 78.2% in 2025, as costs grew 3% and revenue only 1.5%. KPMG expects the number of banks to fall below 70 by 2030. Most banks know they need to change, and most have a plan. The money isn't the problem. The order is. Most transformation plans I see have the same five pillars: operating model, data, AI, technology and skills. Each gets a sponsor, a budget and a steering committee. Too often all five start in the same quarter. And all five compete for the same scarce people: the ten or fifteen operations experts who actually know how the work gets done. The result is predictable, and AI shows it most clearly: four in five banks already use it somewhere, yet 76% expect it to save 2% or less over the next 12 months. AI pilots run on processes nobody has redesigned, so they automate the workarounds. Data teams clean fields no process uses. The core system upgrade gets specified before anyone agrees the target operating model, so it gets built twice. And training teaches tools people haven't been asked to use. The order matters more than the budget Here's how I'd sequence the five, and why. The payback ranges are illustrative, for the operations of a mid-sized bank. They show the logic, not a promise.
Impact
Two things stand out. First, the operating model and data pillars fund the rest. Their savings land in year one and pay for AI and modernization. Second, technology is usually negative in year one. That isn't a reason to skip it. It's the reason it comes after the self-funding pillars, not before. How to measure it without fooling yourself Every pillar needs the same baseline: the as-is process, measured in hours, error rate and cycle time. The future state is designed, not hoped for. The gap between the two is the benefit, and Finance signs it off. Two rules keep the numbers honest. Count each hour once: if AI and training both claim the same saving, one of them is wrong. And don't book avoided fines as ROI. Report risk reduction on its own line. A test for Monday Before you fund the next AI use case, ask four questions: 1. Has the process it touches been redesigned in the last 12 months? If not, start there. 2. Are the errors in its input data under control? If not, fix the root cause first. 3. Who owns the benefit, by name, and has Finance agreed the baseline? 4. Will the people doing the work be trained on the new process, not just the tool? Four yeses: fund it. Any no: you're paying to automate today's problems. The banks that close the gap between record assets and real efficiency won't be the ones spending most on transformation. They'll be the ones doing things in the right order. Source: KPMG Switzerland and University of St. Gallen, Clarity on Swiss Private Banks, 24 June 2026.