Nearly four in five banks (78%) are exploring agentic AI, or autonomous systems capable of executing complex tasks with minimal human input, according to the International Data Corp. One of the main drivers is the push for hyper-personalisation at scale, a goal that challenges banking's long-standing resource model.
Traditional banking follows an 80/20 rule, where 80% of resources serve just 20% of core users who are typically high-net-worth individuals. However, mass-market customers now expect highly customised services, too.
“In the future, everyone will be a VIP customer. They will also be served by a super assistant (or AI agent) via their mobile phone,” says Jason Cao, CEO of Huawei Digital Finance BU, explaining how the service model is fundamentally shifting. “This means that apart from pleasing their customers, banks will also need to satisfy the needs of each customer’s AI agents.”
Reaching that future state, or what Cao terms an "AI bank", requires structural transformation beyond isolated AI experiments. AI-driven change must extend across financial service models, collaboration frameworks, risk decisioning and infrastructure.
Paths to becoming an AI bank
Cao outlines two strategic paths financial institutions are taking to become AI banks. Large organisations tend to build an AI platform first and then deploy applications on top of it. Meanwhile, smaller institutions start with a high-value use case — such as re-engineering an end-to-end lending workflow — before expanding AI to other processes or business units.
Scaling those efforts is where most banks struggle. “Scaling an AI experiment from the lab to across an organisation means confronting engineering issues such as data quality and accessibility, infrastructure bottlenecks, integration problems and model drift. The reality is that most banks aren’t accustomed to managing these or don’t have the knowledge or talent to address all those challenges,” says Cao.
IT architecture decisions compound the challenge. Cao warns against incremental system additions that create technical debt. "Incremental system changes could lead to a complex, unmanageable IT architecture. That's why we believe in taking an evolutionary approach with a target IT architecture in mind.” Banks must also ensure interoperability between legacy systems and AI-native applications while maintaining a total view of enterprise-level architecture, he adds.
Huawei's full-stack approach
Huawei is positioning itself to help banks accelerate this shift through a four-layer approach covering scenario intelligence, AI platforms, data and knowledge platforms, and underlying infrastructure. It has already supported global financial institutions in deploying more than 500 AI applications, including intelligent credit, operations, risk control and marketing.
At the core sits Huawei's intelligent computing platform, designed to handle the high concurrency and low latency demands of AI inference at scale. This computing foundation supports enterprise-wide AI data governance through a unified knowledge and data platform, integrated with financial agent platforms and data and model engineering practices to create an end-to-end loop of AI capabilities.
The approach has delivered tangible results. For example, Huawei and a major Chinese bank jointly developed a next-generation intelligent mobile service architecture. Built on intelligent computing infrastructure and AI platforms, the system leverages hierarchical multi-agent collaboration, long-term memory storage, and end-to-end performance optimisation across hardware and software. The deployment achieved an intent recognition accuracy of above 90% and a latency of as low as 1.2 seconds, helping the bank shift from a passive response to a proactive service.
Co-creation as the AI accelerator
The complexity of deploying AI has prompted banks to adopt collaborative development or co-creation. Huawei has formalised this approach through its RongHai partner programme, which aims to co-create a "six-capability cluster" with global partners. The cluster systematically builds an AI ecosystem covering model development, agent engineering, industry knowledge bases, and scenario-based applications, while optimising deployment efficiency through end-to-end hardware-software integration.
The approach promotes global replication of industry best practices. It has already proven effective with multiple banks in China and is now being replicated with financial institutions overseas, says Cao. For instance, a digital bank in Southeast Asia worked with Huawei and its partner to develop a new core banking system within 35 days, compared to the six months to a year such projects typically require.
To further accelerate AI adoption in the financial services sector, Huawei recently launched its FinAgent Booster (FAB) platform, which consolidates the company’s engineering expertise into ready-to-use templates, solutions, and toolkits. By optimising models and computing resources, FAB reduces the technical complexity and development time of AI deployments. This enables banks to roll out AI agent applications more quickly and achieve business value faster.
Using FAB’s intelligent loan-review template, Huawei helped a city commercial bank in China launch an automated review system in just two weeks, significantly ahead of the typical months required. The FAB platform also supported a major bank in the Middle East in deploying an intelligent credit card approval system that cuts decision times from 10 minutes to 20 seconds, boosting efficiency and customer experience.
Despite these quick wins, individual use cases represent only the starting point for a comprehensive AI transformation to become an AI bank. “Financial institutions should not underestimate the long-term impact of AI, nor should they overestimate short-term results. AI transformation is not easy, but it is a must. This is why partnership is critical to set the pace and accelerate the transformation of banks. For sure, we try to be such a partner,” says Cao.
