Beyond Digital: How Resilience Empowers AI-Infused Finance

Now, LLMs have become low-cost and small in size, which means AI is no longer exclusive to leading banks. Small and medium-sized banks can leverage AI effectively.

And open-source LLMs allow every bank to have its own model, which resolves three problems:

• industry scenario adaptation
• independent innovation and
• the need to store sensitive data locally

The chain-of-thought technology provides transparent traceability of AI decision-making processes. This meets regulatory and audit requirements, making AI not only a tool for tasks but also a trusted assistant in the core business of banking.

What will the future of banking look like? We believe every bank will have its own AI agent and two types of employees. Each bank will develop its own AI agent. Models and data will become the most critical assets for every bank.

In the future, banks will only need two types of employees: AI employees and AI-powered employees. With the support of AI agents, each employee can handle more complex tasks efficiently.

AI will infuse every aspect of banking, from user devices and channels to marketing, risk control, and R&D. The way banks operate and serve customers will fundamentally change.

AI-powered devices redefine user experience

LLM inference is now embedded in smartphones faster than ever. For example, in mobile apps, AI agents can now handle money transfers in just 2 steps instead of 10. It’s a huge leap for user experience!

AI takes marketing to the next level with hyper-personalized services

AI is transforming financial marketing. With AI, we can precisely target customers, tailor services, and choose the perfect timing for outreach. This creates hyper-personalized experiences. Account managers can now handle over 400 clients each, up from less than 100. This has significantly boosted conversion rates.

AI powers smarter decisions and real-time risk management

AI is completely changing how we handle risk control as well. With its thought chain capability, it can solve the challenge of auditability, making decisions transparent and traceable. Combined with graph engines, AI enables real-time fraud detection, making financial systems safer and more reliable.


AI unlocks opportunities but brings challenges

AI is growing fast, bringing big chances but also big challenges for infrastructure. One of the biggest risks from AI is security. Hackers using AI can make viruses harder to spot, change faster, and attack more often. This makes stopping them much tougher. With AI entering mobile devices, the users of apps will shift from humans to AI agents. This will increase performance demands a thousandfold, impacting wide-area networks, and data centers. AI in risk control needs to work 24/7 to protect real-time transactions. This means AI inference needs dual or multi-active setups, reshaping infrastructure. It puts huge pressure on AI hardware, software, and inter-data center networks.

At last year’s MWC, we set a bold vision for the future of financial infrastructure: the 4 Zeros. These are: Zero Trust Security, Zero Wait Experience, Zero Downtime Availability, and Zero Touch Operation.

This year, we’re introducing the R-A-A-S framework to achieve the 4 Zeros. It includes:

• Reliability
• Availability
• AIOps
• Security architectures
  

According to the design principle, reliability is very important. Its target is zero RPO. How can we approach it?

In terms of active-active data centers, we believe that using three key technologies, banks should design multiple copies for the protection of storage, keep real-time synchronization of data, and the most important factor is multiple technology collaboration among storage, network, and computing.

The second one is availability with 99.999%. How can we achieve it?

First of all, we should pay more attention on cell-based cluster design for databases and microservices. In addition, we should focus on a unified multi-active service platform.

The third one is autonomous operation. We put forward 0-1-5-10 as its target. What’s the meaning of this? It means 0 human errors, 1-minute discovery, 5-minute location, and 10-minute recovery.

Across network-security-cloud and across multiple vendors, we need to build a unified AIOps system to manage the whole infrastructure. By using digital twin technologies to simulate the network to escape the issue in advance and using AI technology with the indicator library, fault mode library, and repair script library, we can quickly troubleshoot and recover faults.

The last one is security. Fast blocking response is very important. It should be reduced from minutes to seconds. Three aspects of technology should be taken into account for E2E security of the whole network.

Collaboration across domains is very important. Using secure endpoints, branch firewalls, and data center firewalls to build terminal–edge-cloud collaborative protection, and in data centers, building cloud-network-storage collaborative protection, then an in-depth security management system is necessary. This can coordinate commands in a unified manner, aggregate information from all defense lines, and deliver security policies to each defense line. Finally, we should emphasize that we also have to use AI to fight AI, detecting viruses to block them in advance.

Huawei provides computing, storage, network, cloud, and AI products and solutions, provides R-A-A-S architecture design, and is willing to help global financial customers build a solid digital foundation, and accelerate customers move forward to AI-infused finance.

Learn more about Huawei intelligent finance solutions.

Disclaimer: Any views and/or opinions expressed in this post by individual authors or contributors are their personal views and/or opinions and do not necessarily reflect the views and/or opinions of Huawei Technologies.