This article is part of the monthly AI Radar series, providing a recap of innovative AI deployments and related company announcements in telecommunications, financial services, and customer experience fields of practice. It also offers insights into companies deploying AI, aimed at assisting business executives and technology leaders develop their own AI projects and long-term strategies.

Banks today are facing a lot of pressure to stay up-to-date and competitive due to economic challenges and the rise of digital banking platforms that are attracting more customers with new propositions. Traditionally, these heavily regulated institutions haven’t been the quickest to adopt new technologies, but now they really need to step up. They have to not only improve their usual services but also bring in new technologies to cut costs and improve the way they serve customers, which helps them stand out from the competition.

One key technology making its way into many banks is AI, particularly Generative AI. This technology has the potential to dramatically increase productivity, tailor customer journeys, and allow banks to concentrate on value creation.

In the past, banks have relied on rule-based tools and basic statistical methods for activities such as monitoring and authorising transactions, which often misses out on new types of fraud and money laundering techniques. But now, banks have come a long way in using AI and machine learning in operations, to scrutinise financial transactions in greater detail. This helps them spot fraud more accurately with fewer mistakes, which makes monitoring transactions more efficient. Banks are also using AI to spot suspicious patterns and connections that might not be obvious to human analysts. For example, most banks today use artificial neural networks to predict illegal actions by bad actors by identifying hidden links between customers and employees, which greatly improves their ability to prevent crimes and protect valuable assets.

Despite these advances, the banking industry is still early in its journey of using AI. They are moving forward carefully because they need to follow strict rules. Banks are watching how other sectors like healthcare and telecom use AI before fully using it themselves. Right now, most banks are not ready to use AI for customer-facing services and it might take at least a year before they do. Bank leaders are setting up test beds and starting pilot projects to evaluate the technology’s capabilities and limitations. Despite an overall cautious industry approach, mainly because of regulatory and security issues, some banks are ahead of others in operationalising AI, slowly starting to use AI across its value chain to capitalise on emerging opportunities.

Here’s what’s new this month in AI within the banking sector:

OCBC deploys array of Gen AI apps

OCBC has introduced a range of Gen AI applications to enhance productivity and streamline operations across its global workforce:

• OCBC GPT: A Gen AI chatbot deployed within OCBC’s private Microsoft Azure environment, integrated into the Microsoft Teams platform used by its 30,000 employees. This tool assists with tasks such as writing investment research reports, translating content, and drafting customer responses, helping OCBC employees complete their tasks about 50% faster than before.
• OCBC Wingman: A coding assistant that provides over 1,000 developers the ability to auto-generate, debug, and improve computer codes. Built from an open-source model and deployed in less than two weeks, Wingman now handles approximately 30,000 requests daily, effectively increasing developer productivity by 20 to 30%. It has been updated based on user feedback to include drop-down menus for common tasks like writing tests.
• OCBC Whisper: This speech-to-text technology is used in OCBC contact centers to transcribe and summarise calls in real-time. It also analyses sales calls to identify potential anomalies in the sales process, contributing to enhanced agent performance and personalised development programs.
• OCBC Buddy: An internal knowledge base chatbot that provides quick access to information on bank policies, medical claims, and annual leave, among other topics. It searches through over 150,000 pages of OCBC’s internal documents and is used around 30,000 times monthly. Additionally, it can record and transcribe face-to-face meetings, immediately emailing the transcription to the meeting owner.
• Document AI: This search tool streamlines the extraction and summarisation of key information from internal documents, reducing the time needed to read each document from 30 minutes to just one minute. It is primarily used by employees handling various financial, risk, and ESG documents.

Mizuho partners with Fujitsu to improve operations with AI

Mizuho Bank is pioneering the use of Gen AI to optimise its system maintenance and development. In partnership with Fujitsu, the bank is leveraging the Fujitsu AI Platform combined with Fujitsu’s engineering expertise to enhance the quality and resilience of its systems. This collaboration focuses on using AI to automatically detect errors and omissions in system design plans and audit processes. Additionally, the initiative aims to develop techniques for automatically generating test specifications and source codes from system design plans using Gen AI.

To further harness the potential of AI, Mizuho has deployed Gen AI tools across its 45,000 staff in core lending and brokerage units nationwide. This rollout includes the Wiz series of applications designed to streamline complex and time-intensive processes:

• Wiz Chat: A chatbot that allows employees to personalise and generate content in response to customer queries. This tool helps staff bypass the traditional need to consult an in-house support desk by providing direct access to over 30,000 pages of administrative procedures, FAQs, and guides.
• Wiz Search: This tool supports employees with business operations queries, enabling them to quickly find and apply detailed procedural information on how to approach complicated customer cases and proceed with service provision.
• Wiz Create: Focused on enhancing the efficiency of preparing credit approval documents, Wiz Create allows sales staff to generate customised approval materials in a single click. By integrating financial statements and interview records, this tool drastically reduces the preparation time from 1-2 hours to approximately 10 minutes per case, while also ensuring a consistent quality standard in the documentation.

Wells Fargo teams up with Google’s PaLM2

Wells Fargo is leveraging Google’s advanced AI technologies to power its virtual assistant, Fargo, aiming to improve customer service and regulatory compliance. The bank utilises Dialogflow, Google’s conversational AI platform, to enable Fargo to handle a wide array of customer inquiries through voice and text in its mobile application. Additionally, Fargo incorporates Google’s PaLM 2, enhancing its ability to clarify what information clients must provide to regulators and assist with everyday banking queries. Since its launch in March 2023, Fargo has managed 20 million interactions with 2.7 interactions per session — on tasks such as paying bills, sending money, and searching transactions — and is expected to achieve an annual interaction rate of 100 million.

Fargo is designed to support a variety of banking functions, such as providing insights into spending patterns, checking credit scores, processing bill payments, and offering detailed transaction information. According to CIO, Chintan Mehta, Fargo is on track to sustain a 100 million interaction run rate per year.

Furthermore, Wells Fargo integrates LLM technology into its Livesync product, which aids customers in goal setting and planning, offering tailored financial advice. This strategic use of Gen AI underscores Wells Fargo’s commitment to enhancing its digital services and customer interaction capabilities through innovative technology. Wells Fargo also deployed other applications that use open-source LLMs, including Meta’s Llama 2 model, for internal uses and has built an AI platform Tachyon for multi-model applications.

ADA: Advancing DBS with AI

At the core of DBS’s AI initiatives is ADA, a self-service data platform that consolidates bank data to enhance discoverability, quality, governance, and security. ALAN, an AI protocol platform, integrates with ADA to enable widespread AI use, significantly reducing the time to develop AI models — from 15 months to as little as two weeks. This integration supports data scientists in rapidly deploying AI solutions to enhance operations and decision-making. Other AI implementations include:

• DBS NAV Planner: This financial and retirement planning tool is equipped with an AI-driven recommendation engine, known as the “Make Your Money Work Harder” digital advisor, that recommends the right investment products based on individual customer profiles covering financial goals, risk tolerance, investment preferences, and current financial situation. Additional safeguards are in place for customers with a ‘zero’ risk profile or negative cash flow, directing them towards safer financial solutions or steps to improve their financial health.
• Jobs Intelligence Maestro (JIM): Partnering with impress.ai, DBS launched a virtual recruiter, revolutionising high-volume hiring processes particularly for high-demand roles like Wealth Planning Managers and Graduate Associates. JIM automates resume reviews, pre-screening queries, and scores online assessments, efficiently shortlisting candidates. It operates as a concierge, guiding candidates through the application across multiple platforms in several languages including English, Bahasa Indonesia, Simplified Chinese, and Traditional Chinese. JIM can answer FAQs with a 97% success rate and redefining the application process to be more seamless and engaging. Since its pilot, JIM has significantly improved recruitment metrics: screening over 10,000 candidates, identifying over 880 successful hires across Asia, reducing hiring time by 75% or 40 hours per month, and lowering candidate attrition from 15% to 3%.

UOB pilots Microsoft Copilot at scale

UOB is the first Singaporean bank to implement Microsoft 365 Copilot across various business functions, involving 300 employees from branches, customer service, technology, and operations. This tool integrates with Microsoft applications like Outlook, Word, PowerPoint, Excel, and Teams, leveraging UOB’s extensive data from documents, emails, chats, and meeting notes to boost productivity, accessibility, and collaboration.

Three key benefits for employees:

• Enhanced Communication: Client-facing employees can use Copilot to craft clear, professional messages tailored to each customer interaction. They can repurpose existing materials to build engaging presentations for meetings with potential customers.
• Greater Productivity: Copilot can easily summarise long documents and emails threads, quickly transform raw data and documents into visualised charts, and routinely deliver concise summaries and action points of meetings in Teams.
• Higher Accessibility: Copilot makes it easy for employees to locate and reference key information within the bank knowledge base, further reducing response time to customers and improving customer satisfaction. It smartly pulls relevant data from various documents or emails based on what the employee is working on, improving overall work efficiency.

Guided by the Monetary Authority of Singapore’s (MAS) FEAT principles — Fairness, Ethics, Accountability, and Transparency — UOB ensures that its AI deployment adheres to ethical standards and regulatory requirements. These principles ensures that AI-driven decisions must be fair, accurate, ethical, accountable, and transparent, aligning with UOB’s commitment to responsible AI usage.

Banks of tomorrow are driven by AI

As Gen AI continues to act as a powerful lever for growth in the banking sector, its role in driving cost reductions and revenue increases cannot be understated. According to Accenture, banks can expect to see a 9% to 12% reduction in mid- and back-office costs due to a 7% to 10% increase in productivity in corporate functions. Furthermore, Gen AI could enhance the time spent on client interactions and advice — which make up about 80% of banking revenue — by 17%, potentially leading to a 9% increase in revenue.

In response, banks are actively recruiting data scientists, engineers, software developers, and other AI specialists, even as they reduce headcount in other areas. Due to the high costs associated with training Gen AI models, many banks are opting to use commercial large language models hosted on public clouds instead of developing their own. This approach not only saves on costs but also mitigates the risk of relying too heavily on a single AI vendor. Diversification is crucial, leading to many banks preferring a multi-cloud, multi-model provider strategy.

To capitalise on AI, banks must transition from a legacy architecture and operating model to an automation and cloud-first strategy. Building the core technology and data capabilities upon a highly automated, hybrid cloud infrastructure is essential to gain competitive and differentiating capabilities. Leaders should focus their transformation on developing robust API strategy and streaming architecture, implementing lightweight core platforms, building modern data and analytics capabilities on cloud infrastructure while ensuring the availability of the right skills and talent required to remain competitive and agile in a market where finding the right talent for AI roles is becoming increasingly challenging.