Deutsche Bank is reducing 15 core banking systems to two, while TD and China Merchants Bank are tying AI to frontline work, credit processing and financial returns. Modernisation is moving deeper into the systems and workflows that shape cost, capacity and customer service.

Expansion is also testing banks’ control environments. The Financial Stability Board (FSB) and Financial Conduct Authority (FCA) found that frontier AI can expose vulnerabilities faster than firms can fix them, while Visa and the US Financial Crimes Enforcement Network (FinCEN) are strengthening how institutions detect and trace scams. Hong Kong and Singapore are supporting deployment through supervised trials and co-funded adoption.

Read more on the week’s key developments

1. Deutsche Bank selects Thought Machine as it consolidates 15 core banking systems

Deutsche Bank selected Thought Machine’s cloud-native Vault Core platform as the core banking engine for Personal Banking and Wealth Management products in Germany. It is the first of two platforms the Private Bank plans to use as it reduces 15 core banking systems. Testing is scheduled for late 2026, followed by phased product migration from 2027 with the existing and new systems operating in parallel. As part of the wider programme, the Private Bank plans to invest around EUR 600 million ($699 million) in technology, operations and AI by 2028 and generate approximately EUR 300 million ($350 million) in annual run-rate savings.

The project carries particular execution weight after Deutsche Bank’s 2023 Postbank technology migration led to severe customer-service problems and regulatory oversight. Vault Core will sit within the bank’s established German franchise, where migration affects existing Personal Banking and Wealth Management clients. GFT has been appointed systems integration partner for what Thought Machine describes as a decade-long technology partnership.

2. TD quantifies AI value as frontline deployment expands

TD, Canada’s second-largest bank by total assets, reported CAD 195 million ($141 million) in year-to-date value from AI, close to its CAD 200 million ($144 million) FY2026 target. Generative AI assistants now support more than 20,000 client-facing employees in Canada, while TD Auto Finance Canada has automated around one-third of its manual funding processes and introduced digital income verification to shorten credit decisions.

TD is placing AI within its medium-term financial plan, targeting approximately CAD 500 million ($361 million) in annualised cost savings and the same amount in annualised revenue uplift. The route to those targets runs through high-frequency activities such as frontline servicing, contact centres, credit assessment and software development. This gives TD one of the more developed bank-wide frameworks for moving AI from employee tools into operating and revenue performance.

3. China Merchants Bank expands AI deployment and cuts SME credit-processing time

China Merchants Bank, one of China’s largest national joint-stock commercial banks with CNY 13.71 trillion ($2.04 trillion) in assets at end-June, reported that it had deployed 256 domain-specific models across 1,386 AI-enabled scenarios. The bank said these applications produced efficiency gains equivalent to 13.88 million employee working hours. In SME lending, AI reduced average credit-service time from 36 hours to 2.72 hours and can generate around 90% of the content in some due-diligence reports.

The deployment spans a corporate franchise serving 3.86 million customers, with corporate loans reaching CNY 3.2 trillion ($476 billion), up 9.27% from the end of 2025. CMB also said employees adopted 68% of AI-generated post-loan monitoring findings, while warnings were triggered 45 days earlier than under conventional processes. This extends its use of AI from preparing documents into credit decisions and ongoing portfolio monitoring.

4. FIS pilots embedded banking platform with M&T, Cogent and Commercial Bank of California

FIS launched an embedded banking platform that allows banks to provide accounts and payments within third-party business software. Accounts remain on the participating bank’s balance sheet instead of being held through a separate third-party ledger. M&T Bank, Cogent Bank and Commercial Bank of California are participating in the initial pilot, with account and payment capabilities planned for the fourth quarter of 2026.

FIS is using its existing banking infrastructure to help financial institutions compete in a market largely developed by fintech banking-as-a-service providers. The initial participants range from a large regional lender in M&T to smaller commercially focused banks, giving FIS a test of the model across different balance-sheet sizes. Banks retain the deposits and regulated accounts, while business-software providers give them access to customers at the point where financial services are needed.

5. Visa integrates Featurespace into account-to-account fraud protection

Visa enhanced its A2A Protect service with a unified fraud score combining its network intelligence with technology from Featurespace, which it acquired in December 2024. Participating banks can use network-level data to identify scam hotspots and coordinated activity across institutions. Visa said the service had increased fraud detection by 75% during the first six months of deployment and reported reductions in fraud and unnecessary alerts at early users.

This is Visa’s first product integration of Featurespace and forms part of a larger expansion into fraud and identity services beyond card payments. In August, it agreed to acquire behavioural-intelligence provider BioCatch for $2.4 billion, adding technology that evaluates user behaviour before a payment is initiated. Featurespace contributes transaction-level detection, while BioCatch would extend Visa’s coverage into account opening, account takeover and scam activity occurring earlier in the customer journey.

6. FSB and FCA raise cyber-resilience concerns as frontier AI accelerates vulnerability discovery

The Financial Stability Board warned G20 finance ministers and central bank governors that frontier AI could increase the speed, scale and sophistication of cyber threats. It called for responsible model deployment, stronger response and recovery capabilities at financial institutions and greater resilience among critical technology providers. A separate Financial Conduct Authority review found that firms are using frontier models to identify and combine vulnerabilities faster than engineering, validation and remediation processes can address them.

Together, the findings place remediation capacity and technology dependencies at the centre of the frontier AI risk discussion. Faster discovery can itself become an operational-resilience problem if firms cannot prioritise findings, deploy patches and manage urgent system changes without disrupting important services. Shared cloud, software and infrastructure providers continue to extend that pressure across institutions.

7. FinCEN adds scam-centre indicators to financial crime reporting guidance

The US Financial Crimes Enforcement Network issued an alert on transactions linked to transnational criminal organisations operating scam centres, particularly in Southeast Asia. It provided behavioural, financial and technical indicators and asked institutions filing related suspicious activity reports to use a dedicated reference term. A separate FinCEN analysis identified 33,904 reports involving approximately $12.7 billion in suspected digital asset investment scam activity between September 2023 and December 2025.

FinCEN found that individual institutions often see only one stage of a scam as funds move through bank accounts, digital asset exchanges and off-platform wallets. Its guidance encourages transaction-monitoring, cybersecurity and blockchain data to be combined within investigations and suspicious activity reports, giving law enforcement a more complete view of funds moving across platforms and institutions.

8. Swiss B2B bank Incore tests agentic AI for KYC assessment

Swiss-regulated B2B transaction bank Incore Bank partnered with Kyndryl, the IT infrastructure company spun out of IBM, and Google Cloud to test Gemini-powered agents for customer onboarding and ongoing KYC reviews. Zurich-based Incore serves banks, financial intermediaries and companies across traditional and digital assets. The system achieved up to 99% accuracy in extracting information from onboarding documents and indicated that some onboarding processes could be shortened from months to days.

The agents gather evidence, identify risk factors and produce explainable risk scores and auditable decision records. For a B2B bank serving financial intermediaries and digital-asset firms, faster onboarding can increase its capacity to add and review institutional clients. The use of policy-based controls and human oversight also makes the project relevant to compliance operations where risk decisions must remain traceable.

9. Hong Kong regulators select 36 trials for cross-sector agentic AI sandbox

The Hong Kong Monetary Authority, Securities and Futures Commission, Insurance Authority and Mandatory Provident Fund Schemes Authority selected 36 agentic AI use cases from 30 financial institutions for the first GenA.I. Sandbox++ cohort. The trials involve 27 technology partners and will use computing resources provided by Cyberport. Participants include Bank of China (Hong Kong), Hang Seng Bank, livi bank, Manulife Hong Kong and China Asset Management (Hong Kong), with technology providers including Google, IBM, Tencent Cloud and Visa.

The original sandbox began with 15 use cases from ten banks in 2024 and expanded to 27 use cases from 20 banks in its second cohort. Sandbox++ extends the framework across banking, securities, asset management, insurance, pensions and stored-value facilities. It also marks a shift from content-generation tools toward agents handling multi-step workflows in risk management, fraud prevention and customer service under the supervision of four financial regulators.

10. MAS commits SGD 220 million to fintech adoption and infrastructure

The Monetary Authority of Singapore committed SGD 220 million ($173 million) over three years under the fourth edition of its Financial Sector Technology and Innovation scheme. The funding spans six tracks covering institutional projects, AI adoption, shared infrastructure, innovation centres, talent and fintech awards. Its AI Pathfinder track will co-fund financial institutions adopting market-ready solutions listed on the PathFin.ai platform.

The scheme expands a programme that has supported more than 350 projects since 2015 and follows the SGD 150 million ($118 million) committed under its previous three-year phase. The larger allocation will support frontier technologies and the infrastructure and talent needed to implement them. AI Pathfinder adds a more direct route from testing to adoption by connecting financial institutions with market-ready providers and co-funding deployment.