The banking giants of China and the United States are embracing generative AI.

Written by: Samora Kariuki

Compiled by: Shenchao TechFlow

Global AI Wave

How do banks actually utilize generative AI?

If we set aside the headlines and hype, the essence of the question is: how are the world's largest banks actually using generative AI? It's not about future potential or vendor promotions, but where are the actual applications that have already been implemented?

In the past two years, the global financial industry has quietly entered the era of generative AI. However, this process is not uniform; rather, it presents a pattern of internal and external differences: the low-key deployment of internal tools, cautious experiments aimed at customers, and a few bold innovations are gradually reshaping the internal structure of the banking industry.

Start from the inside and gradually extend outward.

The application of AI has a common point: starting from internal productivity tools.

The main applications of generative AI are focused on enhancing internal productivity—these tools help employees accomplish more work with fewer resources. From JPMorgan's analyst assistant analyzing equity research to Morgan Stanley's GPT-driven tools supporting wealth management advisors, the early emphasis has been on empowering banking professionals rather than replacing them.

Goldman Sachs is building AI assistants for developers; Citi's AI summary tool helps employees process memos and write emails; Standard Chartered's "SC GPT" has been launched among its 70,000 employees, covering everything from proposal writing to human resources issues.

Given that we are in a highly regulated environment, the deployment of internal tools is particularly reasonable. This allows banks to experiment and enhance AI capabilities without crossing regulatory lines. If we refer to the recent actions of the CBN (Central Bank of Nigeria) against Zap, then "better safe than sorry" is clearly a wiser choice.

Business Line Observation: Where is the Value?

The progress of AI applications varies across different departments. There are differences in the speed at which various business units adopt generative AI. Among them, retail banking is in the lead in terms of transaction volume. In this field, generative AI-driven chatbots like Fargo from Wells Fargo and Erica from Bank of America handle hundreds of millions of interactions each year. In Europe, Commerzbank recently launched its own chatbot, Ava.

However, the problem is that some of these tools do not actually use generative AI, but rely on traditional machine learning techniques. For example, Bank of America's Erica works more like a "Mechanical Turk," meaning it creates the illusion of automation through human operation. Nevertheless, what matters is the experiments themselves, rather than the technical labels.

In corporate and investment banking, transformation is more implicit. JPMorgan's in-house tools primarily support research and sales teams, rather than direct-to-customer. Deutsche Bank uses AI to analyze customer communication logs, which are not customer service, but data-enabled, helping bankers understand and serve customers faster and better.

Wealth management sits between the two. Morgan Stanley's AI tools do not directly interact with clients but ensure that advisors are well-prepared before each meeting. Deutsche Bank and First Abu Dhabi Bank are piloting assistants aimed at top clients, designed to answer complex investment questions in real time.

Regional differences: Who's leading the way?

Source: Evident AI Index

North America is leading the way as expected. U.S. banks such as JPMorgan, Capital One, Wells Fargo, Citi, and Royal Bank of Canada (RBC) have turned AI into productivity engines. Thanks to their partnership with OpenAI and Microsoft, they are the first to get access to cutting-edge AI models.

Europe is being more cautious. Banco Bilbao Vizcaya Argentaria (BBVA), Deutsche Bank, and HSBC are conducting internal tests of AI tools and have implemented more security measures. The General Data Protection Regulation (GDPR) has a profound impact on them. As in the past, Europe is more focused on regulation rather than technological advancement, which may come at a cost.

Africa and Latin America are still in the early stages of AI development, but progress is rapid. Brazil's Nubank stands out by collaborating with OpenAI, initially deploying AI tools internally and eventually expanding to customer service. In South Africa, Standard Bank and Nedbank are piloting projects in the AI field, covering risk control, support services, and development.

China: Building an Autonomous AI Technology Stack

Chinese banks are not only using AI but also building AI technology stacks.

The Industrial and Commercial Bank of China (ICBC) has launched "Zhiyong", a large language model with 100 billion parameters, developed in-house. This model has been invoked over one billion times, supporting 200 business scenarios ranging from document analysis to marketing automation. This is not just an application of internal tools, but a fundamental shift in the way the bank operates.

Ant Group has launched two large language models in the financial sector—Zhixiaobao 2.0 and Zhixiaozhu 1.0. The former is aimed at regular users of Alipay, designed to explain financial products; the latter provides support for wealth management advisors, capable of summarizing market reports and generating portfolio insights.

Ping An Group, as a fintech giant integrating insurance, banking, and technology, has gone further. Its developed generative AI assistant AskBob serves both customers and supports account managers. For customers, AskBob can answer investment and insurance questions in natural Chinese; for advisors, it can extract and summarize customer history, product data, and marketing materials, transforming each agent into a digitally enhanced financial expert. Ping An's goal is to redefine financial consulting through AI, not only answering questions but also predicting needs in advance.

In China, the regulatory framework strongly encourages data localization and model transparency, and these institutions have chosen a longer-term path: to build customized AI that can adapt to domestic regulations, language, and market environments. In addition, China has a sufficient talent density, allowing banks to independently develop foundational models, which may be a unique achievement on a global scale.

Who is providing technical support?

Some well-known companies frequently appear globally: Microsoft has become the most common platform through Azure OpenAI. From Morgan Stanley to Standard Chartered Bank, many banks are running their models in Microsoft's secure sandbox environment.

Google's LLM (Large Language Model) is also being used, for example, Wells Fargo utilizing Flan to support its Fargo. In China, it mainly relies on domestic technologies such as DeepSeek and Hunyuan.

Some banks, such as JPMorgan Chase, Industrial and Commercial Bank of China, and Ping An Group, are training their own models. However, most banks are fine-tuning existing models. The key is not to possess the model itself, but to control the data layer and the coordinated operation of the models.

Exploration of the Diversification of AI Applications Worldwide

See the original image in the original text, compiled by: Deep Tide TechFlow

So what?

In a highly regulated industry, it is crucial to act cautiously, which is why banks involve AI rather than standing directly on the front lines. However, as we have seen in other platform transformations, decisive decision-making and rapid experimentation are key. Regulation will never be ahead of execution, and waiting for regulation to be in place before conducting AI experiments is not a wise move. I remember building an agency banking model in a country with no relevant regulation over a decade ago. Once we completed the build, we became the ones explaining this business to the central bank. If I were a member of the bank's board, I would ask, 'How many experiments are we conducting? How much insight are we generating?'

To truly measure progress, we must return to the fundamental principles of platform transformation. Your AI strategy must address the following questions:

"Has our AI strategy rebuilt the core architecture? Has it reduced costs by 100 times? Has it unlocked new value models? Has it stimulated connections within the ecosystem? Has it disrupted the market? Has it achieved the democratization of access?"

The logic is clear - maintaining a skeptical attitude is necessary, but both logic and facts indicate that AI is a new platform revolution. Moreover, logic and facts also show that past platform revolutions have often brought about revolutionary changes in financial markets. For example, Citibank significantly expanded its retail business with technology applications in the 1970s and 1980s. Capital One emerged from nothing to become one of the top ten banks in the market, establishing a significant presence in related industries such as auto loans and mortgages. In Africa, Equity Bank seized the client-server technology wave to become the largest bank by market value in East Africa. Similarly, Access Bank, GT Bank, and Capitec also rode this wave in their respective markets.

The era of AI platforms has arrived, and it will create winners. The focus is not on the losers, but on how the winners capture significant market share in specific areas. For example, Stripe's success in the payment sector is a typical case. These early breakthroughs often lead to market share growth in adjacent fields, such as Nubank becoming an important player in the SME and retail banking sectors through its credit card business.

My point of view is that the winners in the AI era will focus on relationship costs. This is no longer just a transaction game. Transactions have already happened; now it’s a game of customer experience and relationship management. This is the core insight that financial services leaders should focus on. How can we achieve a 100-fold improvement in customer experience and relationship banking at a very low cost? As a bank, how can we leverage intelligent technology to better help customers manage their finances, businesses, and lives? The players who can answer and execute these questions will become the ultimate winners.