UBS pushes AI deeper into operations as "one-bank" model takes shape
UBS is increasing its investments in AI to rework both front- and back-office operations, according to CEO Sergio P. Ermotti on the firm's latest earnings call. The focus: a "one-bank approach" so platforms, data, and process standards travel across businesses instead of living in silos.
"We are investing in a portfolio of large-scale transformational AI programmes designed to increase our operational resilience, enhance the client experience and unlock higher levels of efficiency," Ermotti said. These efforts sit alongside the ongoing Credit Suisse integration.
Where AI hits Operations first
- Client-facing efficiency: triage, recommendations, and service automation to reduce response times and handoffs.
- Middle-/back-office throughput: exception handling, KYC/AML review support, reconciliations, and document processing to cut cycle times.
- Risk and control: anomaly detection, surveillance, and model-driven QA to reduce operational loss and rework.
- Technology stack: standardized data pipelines, model governance, and shared platforms to scale use cases across business units.
Execution at scale
UBS launched more than 300 AI use cases across its operations in 2025 and appointed its first Chief AI Officer, Daniele Magazzeni (formerly at JPMorgan Chase). Ermotti said adoption is "increasingly strong," supported by next-generation tools and platforms that target efficiency and productivity.
The AI buildout runs in parallel with the Credit Suisse integration, which the bank expects to substantially complete by the end of 2026, as reported by Banking Dive. The final wave of client migration is the most complex and is a key dependency to fully winding down legacy infrastructure.
Leadership signals you should pay attention to
UBS's AI agenda is reshaping talent flows. Ronald Jansen, formerly Global Head of the AI Lab in global banking, left for JPMorgan Chase. Mike Dargan, the bank's former Chief Technology and Operations Officer, is now CEO at N26.
Other banks are moving too. Wells Fargo named Saul Van Beurden to lead AI and consumer banking, and the Commonwealth Bank of Australia appointed Ranil Boteju as Chief AI Officer. Competition for AI operators and platform leaders is heating up-expect more reshuffling.
Why this matters for Operations
- Consolidation over fragmentation: a single platform strategy reduces duplicated tooling, data drift, and version chaos.
- End-to-end thinking: design use cases around whole journeys (e.g., onboarding to servicing) to avoid moving the bottleneck downstream.
- Governance as an enabler: standard model approval, monitoring, and controls speed up safe deployment instead of slowing it down.
- Retirement plans for legacy: AI gains won't appear in P&L until old systems are actually turned off-plan sunsetting early.
What Ops leaders should do next
- Prioritize 5-7 high-volume journeys and define the "AI bill of materials" (data, models, prompts, guardrails) for each.
- Stand up a shared feature store and prompt library to reduce duplication and improve reuse across teams.
- Bake controls in: human-in-the-loop checkpoints for material decisions, lineage tracking, and bias/stability testing before scale.
- Create a decommissioning roadmap tied to each use case, with specific systems, interfaces, and manual steps to retire.
- Set operating KPIs: straight-through processing rate, exception rate, average handling time, first-contact resolution, and control incidents.
- Upskill frontline and middle-office teams on AI-assisted workflows using AI Productivity Courses; pair training with new SOPs and performance incentives.
Metrics that prove value
- Cycle time reduction by process (e.g., onboarding, KYC refresh, dispute resolution).
- STP uplift and call/chat deflection without degrading CSAT/NPS.
- Model quality over time: drift alerts, false positive/negative rates, and remediation speed.
- Legacy shutdown milestones hit-licenses, servers, and interfaces actually removed.
Key dependencies and risks
- Client migration timing: late migrations delay legacy shutdown and keep costs high.
- Data quality: poor mapping and lineage kill model performance at scale.
- Change saturation: too many parallel rollouts reduce adoption; sequence in waves and protect critical periods.
- Third-party exposure: clarify where vendor models sit in your control framework and incident playbooks.
If you want a fast scan of practical tools that fit finance operations, this curated list can help: AI tools for finance. For official UBS updates on integration progress and results, see UBS Investor Relations.
Your membership also unlocks:
