AI strategies for banks to match fintech innovation

Customer expectations for banking have shifted permanently. Digital challengers drove the change, setting new standards for speed and simplicity.

Incumbents still have deeper advantages: capital, infrastructure, and customer trust. However, slow systems and siloed data continue to erode these strengths.

AI offers a way to close the gap and turn existing strengths into momentum.

What’s actually happening in the market right now

Banks have capital, infrastructure and customer trust, but lack speed. Fintech companies have speed and usability, but lack scale and regulatory depth. Neither side can fully solve the other’s problem alone.

Banking technology trends across 2025 and into 2026 made several shifts impossible to ignore. Generative AI moved from pilot to production. Legacy systems stopped being merely expensive to maintain. They started actively blocking new capabilities. Regulators began moving faster than most expected.

In EY’s October 2025 survey, eight in 10 banks reported efficiency and productivity gains from AI. Just over half (53%) also reported revenue gains. In Wolf and Company’s 2026 survey, every community bank had adopted AI in some form. Yet only 5% had launched a scaled, governed AI program.

Technology transformation firms working in financial services, including DXC’s financial services division have been frank about a structural problem. Fragmented infrastructure limits what banks can do with their data. This bottleneck stalls AI adoption before it starts. Core banking modernization has shifted from a someday initiative into an operational necessity. The World Economic Forum’s 2026 AI Playbook reaches the same conclusion. It finds firms capture more value when they redesign workflows, organization and technology together.

In 2025, DXC signed a 10-year agreement with Spain’s Unicaja on AI-driven banking operations. The deal is no outlier. Lloyds Banking Group, ING and Standard Chartered are publicly investing in modernizing core systems. Without that groundwork, even strong AI projects end up as cosmetic patches on decades-old logic.

Technologies that moved beyond the lab

Several areas crossed from demos into actual deployment.

Generative AI in customer service: JPMorgan Chase offers IndexGPT, a range of GPT-4-assisted thematic investment baskets. Bank of America has pushed Erica beyond basic chatbot territory toward proactive financial guidance. BBVA has deployed ChatGPT Enterprise across the group, and now offers its app inside ChatGPT.

Agentic AI in the back office: Early deployments handled isolated tasks. Banks are now testing agents that run multi-step processes end to end, from underwriting to compliance. Morgan Stanley’s AI assistant analyzes the firm’s research library for advisors. About 98% of its advisor teams use it.

Real-time payments and embedded finance: FedNow in the US and SCT Inst in Europe provide the rails. Banks are only beginning to build products on top of them. In December 2025, DXC and Aptys Solutions partnered to modernize payments for community banks and credit unions. Aptys serves a network of more than 5,500 institutions.

CoreIgnite and API-first architecture: One sharp tool is converting core banking services into API-accessible components. DXC CoreIgnite does this by overlaying existing cores with a secure, API-driven layer. Thought Machine, whose Vault Core platform works with DXC, makes the potential concrete. The vendor says banks can launch new products in hours rather than months.

Deployments drawing real attention in 2026:

• HSBC is testing AI-driven financial crime detection through Swift’s federated learning pilots, keeping sensitive data inside each institution.
• Citi is piloting agentic AI within its proprietary platform, starting with 5,000 employees. Its assistive tools already reach about 175,000 staff across 80 markets.
• Deutsche Bank served as a design partner for Google Cloud’s Gemini Enterprise for Financial Services.
• Santander and Visa completed Latin America’s first agentic commerce pilot, with AI agents buying goods across five markets.

Why fintech wins on speed, and what banks can do about it

AI trends in banking point to one conclusion. The winning strategy is not to become a fintech. It is to combine structural advantages with the velocity of a technology company.

Approaches that actually work:

• Platform banking: the bank becomes infrastructure that partners build on through APIs. Citi Token Services and JPMorgan’s Kinexys show what this looks like in payments and digital assets.
• Acquisition or partnership: JPMorgan bought the UK digital wealth manager Nutmeg. Santander agreed to acquire TSB in 2025. Visa’s bid for Plaid failed, but even failed deals signal where strategy is pointing.
• Internal innovation hubs: Citi Ventures and ING Labs keep talent engaged and ship real products.
• Technology re-platforming: replace core systems module by module, without a big-bang migration. Lloyds took that approach deliberately and methodically.

Where AI gives banks a real edge over fintech

Most coverage asks how AI helps banks catch up. The better question is where banks hold structural leads fintech cannot copy quickly.

• Credit risk. Banks carry decades of borrower data across multiple economic cycles. Most fintech lenders have not weathered a real recession with their loan books. Models trained on that history tend to be sharper, provided the data is usable.
• Fraud detection. Transaction volume gives banks better training material than any challenger can match. The obstacle is legacy infrastructure that keeps models out of real-time flows.
• Regulatory compliance. Banks can design AI systems with Basel III, DORA and the EU AI Act built in. EY’s research suggests this discipline pays. Banks with formal AI oversight committees report better revenue and cost outcomes from AI.

AI strategies being deployed

From chatbots to personalization that actually works

The first generation of banking chatbots was, plainly, bad. Customers navigated menus for minutes, then heard: connecting you to an agent. Current deployments tell a different story. Banks with LLM-based systems and genuine customer context see better resolution rates.

The key is not the model. It is the data access architecture. A system that knows transaction history, products, and complaints responds at a fundamentally different level. Bank of America’s Erica remains the most cited example. Integration with structured customer data made it work, not the technology itself.

Demand supports the investment. The World Economic Forum (WEF) reports that 71% of banking consumers would welcome an AI assistant. But 82% want to approve the agent’s actions before execution.

Compliance automation and risk management

Compliance may be the most impactful area for large institutions right now. It consumes a disproportionate share of operating budgets. AI addresses that through:

• Automated transaction monitoring for AML workflows
• Regulatory report generation without manual compilation
• AI-assisted review for sensitive operations
• Regulatory text analysis that maps rule changes to internal procedures

HSBC’s experience reflects this trend. Its AI-based anti-money laundering system, built with Google Cloud, cut false positives by about 60%. It also identifies two to four times more suspicious activity than before.

Re-platforming: Modernizing without stopping the train

Institutions that swapped everything at once ran into cost overruns and partial outcomes. The graveyard of stalled big-bang core replacements is well documented.

The standard approach is now incremental. Architects call it the strangler fig pattern:

• New microservices are built alongside the legacy system and exposed through APIs.
• Traffic shifts gradually to the new services.
• Legacy components are decommissioned piece by piece, without one high-risk cutover.

DXC CoreIgnite implements exactly this. Banks unlock functionality progressively through APIs, without betting stability on one migration event.

What’s blocking banks and how the faster ones get around it

• Talent gap. ML engineers with financial services knowledge are rare and expensive. Banks compete with Google, OpenAI and quantitative funds for the same people.
• Data silos. One customer with a deposit account, mortgage and credit card can appear as three entities. AI built on that fragmentation cannot produce coherent customer understanding. Backbase estimates 85% of banking AI projects never reach production. The cause is usually structural, not technical.
• Culture and risk appetite. “We cannot do this because the regulator might object” is a common response. Conservatism that protects depositors can also block low-risk innovation.
• Vendor lock-in. Dependence on a major core vendor makes every change slow and expensive. Negotiating leverage fades quickly after a long-term contract is signed.

Banks moving faster share a common playbook:

• Ring-fence an innovation budget, protected from the core P&L cycle.
• Treat data governance as a prerequisite for AI, not an afterthought.
• Find success islands: ship one product fast without restructuring everything.
• Choose partners who understand financial services constraints.

EY frames the same idea as business-led strategy backed by strong governance.

Where banking technology trends are heading in 2026 and beyond

Scanning banking technology trends for the next couple of years, a few directions stand out as genuinely worth watching:

• Agentic AI in the front office: Agents will execute actions, not just answer questions. Transfers, loan restructuring and account opening become autonomous, with customer confirmation. Accenture calls the end state the 10x bank, where small teams manage fleets of AI workers. The regulatory framework is still forming, which creates a narrow first-mover window.
• Federated learning for cross-bank fraud detection: Banks will not share raw data, but they want to fight fraud collectively. Swift’s pilots with global banks show shared models can work. Sensitive data never leaves each institution’s infrastructure.
• Digital assets and tokenization: JPMorgan’s Kinexys and Citi Token Services are quietly building tokenized asset infrastructure. Pipelines for tokenized bonds, equity and real estate are being laid now. Wolf and Company’s survey shows community banks share this interest. Nearly 40% are prioritizing digital assets and tokenization.
• Quantum computing remains the long horizon: IBM and Google are progressing on portfolio optimization and credit risk modeling. Institutions that skip post-quantum cryptography preparation now will face harder transitions later.

Final thought: Which strategy actually wins

The banks pulling ahead are not the biggest AI spenders. They fixed data and architecture first because AI is the engine. Fragmented data is bad fuel, and no model investment changes the result.

Banking innovation is now an infrastructure race, and the pressure is not easing. Therefore, banks treating technology transformation as a strategic asset are already playing a different game.

Published on April 23, 2026; Updated on September 8, 2026