Banking GenAI #14 - Insights for COOs

Welcome to the latest edition of GenAI for Banking COOs. We hope you are having an excellent start to the year. For those of us in the north-east, it’s a snowy start weather-wise but an intense-start GenAI-wise!

Before we go into our usual banking update, we would like to give a shout to Azeem Azhar and Kai Fu Lee for their wonderful podcast published on Jan 1st on ‘AI in 2025’. You can find it on your podcost platform of choice. We found the conversation fascinating, and it was good to take a step back from the use cases and scaling discussions to where AI is truly going. They touch upon a range of topics such as:

• Agent first design: native AI applications will approach processes of today very differently
• Inference compute: the battle will shift from generic answers from the default LLM to inference analysis that refines the answer; this has material cost and coding skills implications
• Infrastructure design: will become critical to minimize the cost of inference, whether it is the right type of RAM, or how you orchestrate your data. Good old CPU may come back to support the GPU. They even talk about activating small parts of your LLM to get the answer rather than the whole to save costs. Did we use the word fascinating already ??
• Service as software: Moving from software as a service to service as a software as the next generation AI applications and AI agents come online

Hope that got your thinking juices going. Moving to banking, things continue to come fast and furious. In today’s edition, we are covering:?

• Commonwealth Bank of Australia Leads with GenAI to Reduce Fraud and Improve Customer Service
• Mastercard enhances its customer experience by launching a gen AI digital assistantJP Morgan COO Daniel Pinto Speaks on Investment to Equip Workers with Generative AI??

• Banks lead industries in adoption of generative AI technology
• The new “Open AI o1” model series is a drastic improvement to GPT 4’s reasoning capability
• Temenos partners with Nvidia to build on-premises GenAI solution for banks

Happy reading!

?************

Commonwealth Bank of Australia Leads with GenAI to Reduce Fraud and Improve Customer Service

Commonwealth Bank of Australia (CBA) is leveraging generative AI to transform its customer service and operational processes, significantly enhancing both security and efficiency—key priorities for banking COOs. By implementing AI-driven tools like NameCheck, CallerCheck, and CustomerCheck, the bank has achieved a 50% reduction in scam-related losses and a 30% decrease in customer-reported fraud. CBA is also using AI to flag suspicious transactions and send proactive fraud alerts to retail customers, improving safety and customer confidence. In addition, AI-powered messaging systems have reduced call center wait times by 40%, demonstrating the effectiveness of automation in improving the customer experience and operational efficiency.

Beyond fraud reduction, CBA is using Gen AI to streamline services for business customers, such as simplifying loan applications and annual credit reviews. AI-driven solutions can pre-populate loan application documents, speeding up the process and enabling conditional approvals in under 10 minutes. Annual credit reviews, typically taking up to 14 hours, are now expected in just two hours with AI assistance. For COOs, CBA’s approach provides valuable insights into how Gen AI can optimize both customer-facing and back-end operations. The bank’s commitment to responsible Gen AI implementation is reflected in its rigorous governance framework and the integration of 11 guardrails to ensure safe usage, making it a model for strategic, secure AI adoption in banking.

************

Mastercard enhances its customer experience by launching a gen AI digital assistant

Mastercard is leveraging generative AI to augment its customer onboarding process through a new digital assistant. This assistant simplifies the onboarding journey by automating routine tasks and answering critical customer questions using a chat-based interface. Leveraging a large language model with Retrieval Augmented Generation (RAG) and fine-tuning, the tool utilizes Mastercard’s existing onboarding documentation as its knowledge base, ensuring accurate information is provided in response to user prompts. Furthermore, the assistant operates under rigorous data stewardship standards, employing a human-in-the-loop approach to integrate feedback from subject matter experts. This continuous learning model ensures accuracy and up-to-date knowledge, making the onboarding process more efficient and customer-centric.

For a banking CIO, Mastercard’s use case highlights how generative AI can be strategically implemented to enhance existing processes, specifically in customer service. This business case exemplifies a successful partnership with a third-party provider to develop an in-house solution, since the gen AI assistant is the first application released by Mastercard in a collaborative effort with Databricks –a partnership that leverage’s Databricks’ expertise to create a robust AI platform. The example also underscores the importance of responsible and strategic implementation, ensuring that AI tools operate under stringent data governance frameworks to mitigate risks.

************

Banks lead industries in adoption of generative AI technology

Over at ITBrief, Shannon Williams covered a recent report revealing that the banking sector is leading the charge in adopting Gen AI and outpacing other industries. The survey, conducted by SAS and Coleman Parkes Research, shows that 60% of banking leaders have already integrated Gen AI into their daily operations, with most others planning to do so within 2 years. Despite economic uncertainties, banks are dedicating significant budgets to this technology, viewing it as a transformative force.

Banks are applying Gen AI across various functions, with marketing leading the way (47%). Key benefits include enhanced employee satisfaction, improved risk management (cited by nearly 90% of respondents), and greater operational efficiency. Marketing teams are particularly using Gen AI for customer engagement and content creation, with plans to expand into audience targeting and trend analysis.

Challenges remain, particularly around data privacy and governance, with over 70% of banking leaders expressing concerns. While early adopters report significant gains, only 6% of institutions have robust governance frameworks in place. Experts stress that noteworthy AI requires ethical and transparent practices, while others highlight Gen AI’s dual nature as a tool for innovation and a potential risk, particularly in combating fraud.

Editorial comment: The good news from our perspective is that the methodology for Responsible AI has matured significantly in the last year as banks have taken GenAI use cases to production. We have even cracked the code on model risk management. For banks in the US, we have developed and successfully implemented a SR11-7 aligned method that allows models to get over the approval hurdle from model risk management and get stuff in Production! This was successfully used by a leading bank even for customer-facing call-center use cases. Please reach out if you would like to know more.

************

The new “Open AI o1” model series is a drastic improvement to GPT 4’s reasoning capabilities

Talking about inference compute, the models continue to evolve which will further bring more complex banking uses cases into the realm of possibility. Open AI’s new “o1” model series is a significant advancement in artificial intelligence, excelling in reasoning and solving complex tasks beyond its predecessors. The o1 model mimics the thought processes of PhD students in subjects like physics, genetics, math, economics, among others. Specific use cases where the o1 model excels include in-depth research, improved relational reasoning, visual modeling, and complex development & coding work –areas where the GPT-4 model struggled. Open AI also introduced the “o1-mini,” a lean version optimized for speed and cost-efficiency, particularly in generating and debugging complex code. These developments have widespread applications within operations and should lead to increased efficiencies across use cases we’ve previously covered.

The model's rigorous thinking process enhances its effectiveness and safety. This is demonstrated by how there is significant improvement in difficulty to bypass safety measures of the new model – you can further read about this on Open AI’s system card and research post.

The o1 and o1-mini models present opportunities to revisit existing AI use cases, in order to identify opportunities of further increasing the productivity resources already derive from AI tools. The o1 model’s improvements in reasoning and syntax can drive productivity in research and programming tasks as well as other tasks across operations. We are getting closer to the age of AI Agents for Operations.

************

Temenos partners with Nvidia to build on-premises GenAI solution for banks

The convergence of banking and technology continues to accelerate, with artificial intelligence (AI) emerging as a powerful tool to transform the financial landscape. The Banking and Fintech team over at Paypers covered a recent partnership between Temenos and NVIDIA that underscores this trend.

By combining Temenos' expertise in banking software with NVIDIA's advanced AI and accelerated computing capabilities, the two companies aim to empower financial institutions with cutting-edge AI solutions. This collaboration enables banks to leverage the power of generative AI to derive real-time insights from their data, enhancing decision-making and operational efficiency. A key aspect of this partnership is the focus on on-premises AI solutions. By deploying AI capabilities in-house, banks can maintain greater control over their sensitive data and ensure compliance with stringent data privacy and security regulations. Additionally, the integration of explainable AI into critical applications such as credit scoring and anti-money laundering (AML) provides transparency and accountability, addressing regulatory concerns and fostering trust.

This example reinforces an increasing trend of product vendors embedding GenAI capabilities into their solutions. The technology architecture landscape in the future is going to be interesting, where banks will need to decide which capabilities come from which tool, and what is truly needed to be built custom in-house to stitch their various products together. For Operations, given they touch most of the systems in the bank, this is an especially interesting topic to stay close to with their Tech counterparts.

************

We hope you enjoyed reading this edition of the newsletter. We always welcome thoughts and ideas on how we can continue to improve the newsletter and make it more useful for you. Please send them our way. Also, if you have come across relevant content in the banking or AI space that you feel we should cover in this newsletter, please let us know. We are always hungry to learn about new things! See you next time!

************

#GenAI #COO #BankingOps #Innovation #ArtificialIntelligence #BankingTechnology #Banking #Transformation #Operations #ZeroOps #CapitalMarkets #Banking #GenerativeAI #LLMs #ExponentialView #AzeemAzhar #KaiFuLee

This issue of the newsletter is brought to you by Accenture’s Rajat Dev | Nazat Dowla | Francesco Erriquez | Dhruv Laroia | George Heathcote | Alfonso Berguido | Felix Lasnier

For additional details on topics above, please reach out to below:

Editors: Rajat Dev, Dhruv Laroia, Nazat Dowla, Francesco Erriquez

1.?Commonwealth Bank of Australia Leads with GenAI to Reduce Fraud and Improve Customer Service - George Heathcote

2.?Mastercard enhances its customer experience by launching a gen AI digital assistant - Alfonso Berguido

3.?Banks lead industries in adoption of generative AI technology - Felix Lasnier

4.?The new “Open AI o1” model series is a drastic improvement to GPT 4’s reasoning capabilities - Alfonso Berguido

5.?Temenos partners with Nvidia to build on-premises GenAI solution for banks - George Heathcote